✨ feat(server): add per-IP rate limit middleware on /api/v1/greet

Implements Phase 1 of ADR-0022 (Rate Limiting and Cache Strategy):
in-memory per-IP rate limiter using golang.org/x/time/rate. Returns
HTTP 429 with JSON body and Retry-After header when exceeded.

Changes:
- New: pkg/middleware/ratelimit.go (153 lines, 7 unit tests in ratelimit_test.go)
- Modified: pkg/config/config.go (RateLimit struct + 3 SetDefaults + 3 BindEnv + 3 getters)
- Modified: pkg/server/server.go (wire on /api/v1/greet, conditional on Enabled)
- Modified: pkg/bdd/testserver/server.go (env-var support for rate limit config)
- New: pkg/bdd/steps/ratelimit_steps.go (step definitions)
- Added: features/greet/greet.feature scenario (currently @skip @bdd-deferred — see note below)

Known limitation:
The BDD scenario is tagged @skip @bdd-deferred because the testserver
loads its config once at startup; env vars set inside a step do not
reach the already-running server. The middleware itself is fully
covered by unit tests. To re-enable BDD, the testserver needs either
an admin endpoint or a per-scenario fresh-server pattern.

Closes #13 (Phase 1 only — Phase 2 Redis + cache service deferred).

Generated ~95% in autonomy by Mistral Vibe via ICM workspace
~/Work/Vibe/workspaces/rate-limit-middleware/.
Trainer (Claude) finalized the commit/PR step (Mistral hit max-turns).

🤖 Co-Authored-By: Mistral Vibe (devstral-2 / mistral-medium-3.5)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

go.mod

@@ -22,6 +22,7 @@ require (
 	go.opentelemetry.io/otel/sdk v1.43.0
 	go.opentelemetry.io/otel/trace v1.43.0
 	golang.org/x/crypto v0.49.0
+	golang.org/x/time v0.15.0
 	gorm.io/driver/postgres v1.6.0
 	gorm.io/driver/sqlite v1.6.0
 	gorm.io/gorm v1.31.1