Backend chat observability: standardized request summary logging

[akgohain]

Motivation

• Standardize a compact, machine-parseable summary log for chat/workflow-agent endpoints to surface request_id, endpoint, latency_ms, status, and error_type for quick diagnosis of stalls and failures.
• Preserve existing verbose/debug logging while preventing sensitive request payloads from appearing in the single-line summaries.

Description

• Added a dedicated logger and helper function _log_chat_request_summary in server_api/main.py that emits single-line structured summary logs with request_id, endpoint, latency_ms, status, and error_type.
• Instrumented chat endpoints (/chat/query, /chat/clear, /chat/status, /chat/helper/query, /chat/helper/clear) to generate a request_id, compute latency using time.perf_counter(), and call the summary helper on both success and error paths without including request payloads.
• Kept existing endpoint logic and verbose prints/behavior intact so there is no breaking API change.
• Added tests/test_chat_logging_fields.py to assert presence of the standardized fields in the summary log lines and to ensure sensitive text (e.g., the query payload) is not leaked.

Testing

• Attempted uv run pytest -q tests/test_chat_logging_fields.py which failed in this environment due to editable package metadata resolution for pytorch_connectomics (environment-specific issue).
• Ran PYTHONPATH=. pytest -q tests/test_chat_logging_fields.py which passed (3 passed).
• The tests exercise success and failure paths and validate the summary fields and absence of sensitive payload content in the summary logs.

---

Codex Task

[akgohain]

Consolidated into draft PR #165 (feat/codex-wave-integration). Closing this source draft to keep review in one place.