Proposal for a test/README.md

[jeffhhk]

Description

What this PR does:

This change attempts three things!

1. To capture various mechanical implementation details of automated testing in the charmonator repository.
2. To describe when we decide to test manually.
3. To attempt to give a rationale.

Consider all three as both guidance for contributing humans, and also as LLM-guidance for asking LLMs to work with automated tests, similar in spirit to the llms-full.txt guidance created by the MCP community.

Why this change is needed:

There's basically no documentation so far on automated testing. There has been some recent confusion. There are also some improvements in progress, and we should have something to think through a consistent approach.

How to verify/test:

test/README.txt is self-describing.

Related issues/design docs:

• Chat json schema #15
• Feature/tool refactor #16
• Make LLM tests requiring provider keys run in Github Actions #18
• Report V&V metadata #19

Maturity:

The "Scope" and "Automated vs manual testing" sections are preliminary and need feedback. The "Automated testing - Conventions" are more of a statement of what is already implied elsewhere, so may be appropriate to merging immediately and fixing issues later.

Feedback requested:

I would be especially interested in feedback on the "Scope" and "Automated vs manual testing" sections, since they are the most subjective.

Reviewers requested:

Matt, please review.