Per-ESP send-rate limiting (e.g. SES max send rate) #51

Background

The sequence worker runs with concurrency: 10 (mail/sequence-worker.ts:34), and transports are built with no rate limit (mail/transport.ts:166). Nothing throttles sends to an ESP's allowed rate.

Why it matters: Amazon SES enforces a per-account max send rate (often 14/s in new accounts) and rejects anything above it with Throttling. Other ESPs have similar limits. Large broadcasts then fail or churn through retries, and platforms that share one ESP account across tenants need a shared limit.

Proposed fix

• Add a configurable msgs/sec rate per ESP config, defaulting from the provider (for SES, read GetSendQuota → MaxSendRate).
• Enforce it across workers with a BullMQ limiter or a Redis token bucket.
• Back off and retry on throttling errors.

Related: #5 (warmup handles daily volume ramp-up, not per-second rate).

Acceptance criteria

• Sends per ESP config never exceed the configured rate across all workers.
• Throttling errors are retried, not marked failed.
• A test covers the limiter.