Squad

* frontend:
* retitle test results page to "All test results"
* hide non-failing test results by default
* improve formatting of "test results" tables
* link test results to test history in "all test results" page
* ci/lava: don't crash when lava listener provides no status

**Upgrade notes:** this version drops support for starting the celery-related
daemons (worker and scheduler) using the "./manage.py | squad-admin" command
line interface. To start those daemons, you should now use the standard celery
command line interface, i.e.

* `celery -A squad worker` for the worker, and
* `celery -A squad beat` for the scheduler.


**Changes:**

* api:
* Added a REST API for accessing most of the data in SQUAD. This API is only
adequate for read-only access for now.
* The API is self-describing and features a API browser when accessed with a
web browser. This API browser is linked from the navigation bar at the top.
* frontend:
* make login link redirect to original page
* settings:
* respect proxy headers when generating absolute URLs
* upgrade to celery v4
* Drop support for SQL-backed periodic task scheduling
* requirements.txt: drop restriction for using celery v3

* core/notification: don't spam admins

* ci:
* add created, submitted and fetched dates to TestJob
* fill in submission and fetch dates
* admin: avoid loading all Build and TestRun objects
* core/notification:
* fix tests for `notification_timeout`
* core/notification: notify only once on timeout
* frontend:
* display testjob.failure if available

* squad.run: exec() into an actual gunicorn process
* ci/admin: make Job ID a link to the backend
* ci/lava/listener: log only messages related to test jobs of interest
* ci:
* save `TestJob.last_fetch_attempt` before fetching
* delay updating project status until after TestJob is saved
* core
* tasks: make database transaction explicit
* make updating project status optional
* fix failed testjob notification templates

* Revert "squad.run: exec an actual gunicorn process"