Performance harden call drilldown diagnostics

[douglasmonsky]

Summary

• hardens the experimental call drilldown dashboard for large histories
• moves call-origin and dashboard helper metadata into aggregate indexing paths
• removes low-value call/thread anchors and keeps raw evidence on explicit context loading only
• adds SQL-backed live API slices, materialized thread summaries, and incremental source-file refresh metadata
• updates docs, privacy notes, benchmarks, and checklist evidence

Validation

• .venv/bin/python -m ruff check .
• .venv/bin/python -m mypy
• .venv/bin/python -m pytest -q
• .venv/bin/python -m pytest --cov=codex_usage_tracker --cov-report=term-missing
• .venv/bin/python -m compileall src
• dashboard node --check commands
• .venv/bin/python scripts/check_release.py
• .venv/bin/python -m build
• .venv/bin/python -m twine check dist/*
• .venv/bin/python scripts/check_release.py --dist
• .venv/bin/python scripts/smoke_installed_package.py
• .venv/bin/python scripts/smoke_installed_package.py --docker
• .venv/bin/python scripts/benchmark_synthetic_history.py --rows 10000 100000 --json --enforce-thresholds
• .venv/bin/python scripts/benchmark_synthetic_history.py --rows 1000 --with-source-logs --json --enforce-thresholds

Notes

• Normal dashboard_payload and /api/usage remain aggregate-only and must not open source JSONL logs.
• /api/context remains explicit/on-demand for one selected call.
• Evidence caching remains intentionally deferred.