I propose we retire the TestRunner and instead use pytest (or nox*) directly.

As far as I can tell, originally the TestRunner was implemented such that it was called when you run python setup.py test. It was then moved to a new command: python -m iris.tests.runner and we then used that to run the tests in the CI or locally.

Since adopting nox and pytest, we now have multiple layers in how we test Iris:
nox calls the test runner which calls pytest
This seems unnecessary and potentially confusing to a new developer; it's hard to tell where to look for the testing logic. Interestingly our docs only mention running the tests with pytest and nox. I can see the benefit of the test runner as it provides conveniences for running certain tests, but I suspect it may make more sense for that logic to live in pytest (in pytest.ini) or perhaps in nox instead.

So unless there are any objections, I propose we remove the TestRunner.

*Note I have mentioned nox in this issue but considering the move to tox it might be that we use that instead.