test: cover mem_feedback basics

[snoopuppy582]

Fixes #1696

Summary

• Added a dedicated tests/mem_feedback package.
• Covered the base feedback interface contract and MemFeedbackConfig behavior.
• Added dependency-free utility tests for token estimation, update filtering, JSON extraction, chunking, and memory item creation.

Tests

• uv run --python 3.12 --extra mem-reader --with pytest pytest tests/mem_feedback -q
• uv run --python 3.12 --extra mem-reader --with pytest pytest tests/mem_feedback tests/mem_chat/test_base.py tests/mem_cube/test_base.py -q
• uv run --python 3.12 --extra mem-reader --with ruff ruff check tests/mem_feedback
• git diff --check

Notes

• Kept the tests local-only; no LLM, database, or network service is required.
• The contributing guide mentions a dev base branch, but this repository currently exposes main as the active/default base branch.

[CaralHsi]

Hi！
Thanks for your contribution!
This is a good start for basic coverage, but it does not yet cover the main MemFeedback.process_feedback workflow. The PR currently tests config parsing, the abstract base class, and a few utility functions, while the core feedback processing path (_validate_feedback_data, _record_feedback, _analyze_feedback) and operation-specific branches remain untested. Could we add at least one mocked workflow test for process_feedback covering a valid feedback input and the expected downstream calls?

[snoopuppy582]

Thanks for the review. I added a mocked process_feedback workflow test that verifies a valid feedback input calls both _generate_answer and process_feedback_core, then returns the downstream memory record.

Verified locally:

• uv run --with pytest pytest tests/mem_feedback -q -> 16 passed

[CaralHsi]

Looks good to me 😸