feat(governance): per-token query rate limiting & quotas

[xe-nvdk]

Summary

• Per-token sliding window rate limiting (requests/min, requests/hour) with HTTP 429 + Retry-After header
• Per-token query quotas (max queries/hour, max queries/day) with automatic hourly/daily reset
• Per-token max rows per query — queries stop early and return partial results
• Per-token timeout override — enforce shorter timeouts for specific tokens
• Default policies via [governance] config — apply limits to all tokens without explicit policies
• Management API at /api/v1/governance/ — CRUD policies + usage inspection (admin-only, license-gated)
• 3 new Prometheus metrics: governance_rate_limited_total, governance_quota_exhausted_total, governance_policies_active
• License-gated behind FeatureQueryGovernance (query_governance)
• Also untracks docs/progress/ files per updated project instructions

Test plan

• go build ./cmd/... ./internal/... passes
• go test ./internal/governance/... -v — 28/28 tests pass (sliding window, quota tracker, manager CRUD, enforcement, persistence)
• Manual: start Arc with governance enabled + license, create policy with rate_limit_per_minute=5, fire 6 queries, verify 6th returns 429
• Verify governance API returns 403 without valid enterprise license
• Verify queries execute normally when governance is disabled