Week 5 Day 2: add provider aggregate metrics

[utkarshp845]

Summary

• expose process-local Prometheus metrics from the AI SRE Assistant
• aggregate provider outcomes, request latency, deterministic fallbacks, and reported input/output tokens
• keep labels bounded and sensitive prompts, evidence, endpoints, identifiers, and error text out of metrics
• mark Week 5 Day 2 complete in the roadmap, telemetry contract, and build log

Validation

• python -m ruff check app cli tests
• python -m pytest -q — 28 passed
• python -m evals.run_evals — 7/7 cases and 35/35 rubric checks passed