Tomodachi

- Fixes the bug which caused almost all dependencies to be optional installs ("extras") if `tomodachi` were installed with `pip`. All previous required dependencies are now again installed by default also when using `pip` installer.

- Lazy loading of dependencies to lower memory footprint and to make services launch quicker as they usually don't use all built-in implementations. Reference services launch noticeable faster now.
- Optimizations and refactoring of middleware for all service function calls of all built-in invokers, saving somewhere around 10-20% on CPU time in average.
- Improvements to awaiting open keep-alive connections when terminating a service for a lower chance of interrupting last second incoming requests over the connection.
- New option: `http.max_keepalive_requests`. An optional number (int) of requests which is allowed for a keep-alive connection. After the specified number of requests has been done, the connection will be closed. A value of `0` or `None` (default) will allow any number of requests over an open keep-alive connection.
- New option: `http.max_keepalive_time`. An optional maximum time in seconds (int) for which keep-alive connections are kept open. If a keep-alive connection has been kept open for more than `http.max_keepalive_time` seconds, the following request will be closed upon returning a response. The feature is not used by default and won't be used if the value is `0` or `None`. A keep-alive connection may otherwise be open unless inactive for more than the keep-alive timeout.
- Improved type hint annotations for invoker decorators.
- Preparations to be able to loosen dependencies and in the future make the related packages into optional extras instead.
- Printed hints (in development) on missing packages that haven't been installed or couldn't be imported and in turn causing fatal errors.

- Added support for `aiohttp` 3.7.x.

- Documentation related updates. External documentation is available at <https://tomodachi.dev/docs>.

- Note: This is a rather large release with a lot of updates. Also, this release includes a lot of improvements to be able to quicker implement features for the future and modernizes a lot of the build, testing and linting steps to be on par with cutting edge Python development.
- `tomodachi.aws_sns_sqs` and `tomodachi.amqp` decorators has changed the default value of the `competing` keyword-argument to `True`. Note that this is a change in default behaviour and may be a breaking change if "non-competing" services were used. This change was triggered in an attempt to make the API more clear and use more common default values. It's rare that a non-shared queue would be used for service replicas of the same type in a distributed architecture.
- The `tomodachi.aws_sns_sqs` decorator can now specify a `filter_policy` which will be applied on the SNS subscription (for the specified topic and queue) as the `"FilterPolicy` attribute. This will apply a filter on SNS messages using the chosen "message attributes" and/or their values specified in the filter. Example: A filter policy value of `{"event": ["order_paid"], "currency": ["EUR", "USD"]}` would set up the SNS subscription to receive messages on the topic only where the message attribute `"event"` is `"order_paid"` and the `"currency"` value is either `"EUR"` or `"USD"`. If `filter_policy` is not specified as an argument, the queue will receive messages on the topic as per already specified if using an existing subscription, or receive all messages on the topic if a new subscription is set up (default). Changing the `filter_policy` on an existing subscription may take several minutes to propagate. Read more about the filter policy format on AWS, since it doesn't follow the same pattern as specifying message attributes. <https://docs.aws.amazon.com/sns/latest/dg/sns-subscription-filter-policies.html>
- Related to the above mentioned filter policy, the `aws_sns_sqs_publish` function has also been updated with the possibility to specify said "message attributes" using the `message_attributes` keyword argument. Values should be specified as a simple `dict` with keys and values. Example: `{"event": "order_paid", "paid_amount": 100, "currency": "EUR"}`.
- The event loop that the process will execute on can now be specified on startup using `--loop [auto|asyncio|uvloop]`, currently the `auto` (or `default`) value will use Python's builtin `asyncio` event loop.
- Fixes a bug that could cause a termination signal to stop the service in the middle of processing a message received via AWS SQS. The service will now await currently executing tasks before finally shutting down.
- Added SSL and virtualhost settings to AMQP transport, as well as additional configuration options which can be passed via `options.amqp.virtualhost`, `options.amqp.ssl` and `options.amqp.heartbeat`. (github: **xdmiodz**)
- HTTP server functionality, which is based on `aiohttp`, can now be configured to allow keep-alive connections by specifying the `options.http.keepalive_timeout` config value.
- Service termination for HTTP based services will now correctly await started tasks from clients that has disconnected before receiving the response.
- Functions decorated with `tomodachi.aws_sns_sqs` will now be called with the `queue_url`, `receipt_handle` and `message_attributes` keyword arguments if specified in the function signature. These can be used to update the visibility timeouts, among other things.
- The `message_protocol` value that can be specified on service classes has been renamed to `message_envelope` and the two example implementations `JsonBase` and `ProtobufBase` has been moved from `tomodachi.protocol` to `tomodachi.envelope`. The previous imports and service attribute is deprecated, but can still be used. Likewise the optional `message_protocol` keyword argument passed to `tomodachi.aws_sns_sqs`, `tomodachi.amqp`, `aws_sns_sqs_publish`, `amqp_publish` is renamed to `message_envelope`.
- The argument to specify `message_envelope` on the `tomodachi.aws_sns_sqs` and `tomodachi.amqp` decorators is now keyword only.
- The arguments to specify `message_envelope` and `topic_prefix` to `aws_sns_sqs_publish` is now keyword only.
- The arguments to specify `message_envelope` and `routing_key_prefix` to `amqp_publish` is now keyword only.
- `uvloop` is now an optional installation.
- More verbose output when waiting for active tasks during termination.
- Added `tomodachi.get_execution_context()` that holds metadata about the service execution that can be used for debugging purposes or be sent to application monitoring platforms such as Sentry or to be included in custom log output for log search indexing. The `tomodachi.get_execution_context()` function returns a `dict` with installed package versions of some key dependencies, function call counters of different types, etc.
- Refactoring and updates to code formatting, now using Black code style.
- Updated startup output with additional information about the running process, including versions, etc.
- Overall updated documentation and improved examples around running services within Docker.
- `requirements.txt` is no more and has been replaced with `pyproject.toml` with a Poetry section together with the `poetry.lock`.
- Replaced Travis CI with GitHub actions.
- Replaced py-up with GitHub's dependabot, which as of recently also supports Poetry's lock files.
- Added support for `aiohttp` 3.6.x.
- Added support for `aiobotocore` 1.x.x.
- Added `aiodns` as an optional installation, as it's recommended for running DNS resolution on the event loop when using `aiohttp`.
- Updated classifiers for support of Python 3.9.
- Dropped support for Python 3.6.
- The service class decorator `tomodachi.service` is now considered deprecated and the service classes should inherit from the `tomodachi.Service` class instead. This also works better with type-hinting, which currently cannot handle decorators that modify a class.
- The `name` attribute is no longer required on the service classes and if not specified the value will now instead default to `"service"`.

- Changed the order of when to execute the service's own `_stop_service()` function, to always run after active HTTP requests has finished executing, as well as awaiting ongoing AMQP before finally running the user defined function.