Create base annotator test for annotator runner

[bkorycki]

This creates a new base class InteractionAnnotatorTest. It's main distinctions from PromptResponseTest:

• A test must specify an AnnotationType. Unlike regular tests which interface with a list of text completions SUTResponse, tests for annotators expect a certain output from an Annotator. We can build a mechanism later to check for test-Annotator compatibility in the runner.
• It produces test items that model a single-turn SUT interaction (i.e. a prompt + SUT completion pair). This is better aligned with the input structure of annotator objects. Plus, I believe it makes more sense to let Annotators decide how to format their own request, instead of combining a prompt + completion into a single TestItem in the test itself.

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅