Slurry

* Switch to github actions for CI.
* Minor documentation edits.

* Thorough api documentation rework.

(Not released.)

* Introducing external environments. Via the new AsyncSection and SyncSection base classes, it is now
possible to create custom environments to run sections in. In v0.7.0 and v0.8.0, ThreadSection and
ProcessSection was introduced. The environments that ran these sections was hardcoded into the Pipeline
machinery. With this change, the Section itself is responsible for setting up the environment it runs in.
This means, that external sections can now create and manage their environments, and data should flow
between those environments and the main pipeline transparently.
* Api change: Introduces the refine method. The actual work that the pipeline section performs should be
done here.

* New PipelineSection concept. The pipeline section generalizes the types of input that is allowed in pipelines and select sections that take input. PipelineSection objects can either be asynchronous iterables, Sections or Tuples representing sequences of pipeline sections. Using tuples of pipeline sections, it possible to compose nested pipelines to any depth.
* Changed api. Section pumps now take a callable that returns an awaitable as output, instead of an explicit trio.MemorySendChannel. This brings the api in line with synchronous sections, however this is obviously a breaking change.

* New ProcessSection abc. Code can now be spawned in a seperate python process, making true concurrency possible!
* This is a major refactoring. Lots of files have been moved around. Things will break if you upgrade to this version. More breaking changes are planned for 0.9.0 in order to simplify the api across section types.

* New ThreadSection abc. Synchronous python code can now be run transparently next to asynchronous Trio code, using the new ThreadSection class. The pump api is identical to the async version, except that input is a synchronous iterator and the output send function is synchronous function.