perf(api): cache ProjectAuth in Redis and increase DB connection pool

[GenerQAQ]

Why we need this PR?

Benchmark run with 50 concurrent sessions caused cascading API failures: 88 read timeouts + 8 gateway 504s. Root cause: the ProjectAuth middleware hits the database on every request with no caching, which saturated the DB connection pool (maxOpen: 50) under high concurrency. Even though the projects table only has 308 rows with a unique index, queries took up to 1.6s because they were waiting for a free DB connection, not for the query itself.

Describe your solution

1. Redis cache for auth lookups — ProjectAuth now checks Redis first (project:auth:{hmac}, TTL 5min) before falling back to DB. On Redis errors, it gracefully degrades to the DB path. This eliminates the per-request DB hit for authenticated requests.

2. Increased DB connection pool — maxOpen: 50→100, maxIdle: 25→50 to provide more headroom under concurrent load.

Implementation Tasks

• Add lookupProject() helper with Redis-first, DB-fallback pattern
• Update ProjectAuth signature to accept *redis.Client
• Update RouterDeps to include Redis *redis.Client
• Pass Redis client from main.go into router
• Increase maxOpen and maxIdle in config.yaml
• Verify build and tests pass

Impact Areas

• API Server

Checklist

• Open your pull request against the dev branch.
• All tests pass in available continuous integration systems (e.g., GitHub Actions).
• Tests are added or modified as needed to cover code changes.