Add workflow metrics endpoint with analytics and tests

[akgohain]

Motivation

• Provide a backend endpoint to expose workflow-stage and agent-action analytics for observability and downstream UI/consumers.
• Produce aggregate metrics (event counts by type, approval/rejection counts and rates, stage transition counts, and first/last event timestamps) from workflow event histories.
• Ensure the endpoint returns safe defaults when workflows have sparse or no events so callers get a stable schema.

Description

• Added a new router GET /api/workflows/{workflow_id}/metrics in server_api/workflow/router.py with an in-memory seeding helper seed_workflow_events for deterministic tests.
• Implemented compute_workflow_metrics in server_api/workflow/metrics.py to aggregate event type counts, decision metrics (approvals/rejections and rates), stage transition counts, first/last timestamps, and total events while returning safe defaults for empty inputs.
• Exported the router from server_api/workflow/__init__.py so it can be mounted by applications.
• Added test coverage in tests/test_workflow_metrics.py asserting empty-workflow defaults, correctness against seeded events, and a stable response schema for missing workflows.

Testing

• Ran python -m pytest -q tests/test_workflow_metrics.py which completed successfully with 3 passed.
• Attempted uv run pytest -q tests/test_workflow_metrics.py but it failed due to an editable package metadata issue for pytorch_connectomics in this environment (packaging metadata error); this did not affect the targeted unit tests which passed under python -m pytest.
• All new endpoint-focused tests added in tests/test_workflow_metrics.py passed under the working test runner.

---

Codex Task

[akgohain]

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