fix(logging): standardize request-id propagation across services #1487

Request IDs were being generated inconsistently across our services, making it painful to trace a single user request from the web client through the API gateway into the billing worker. Some services used x-correlation-id, others used x-request-id, and the billing worker generated its own.

What changed

• Standardize on the x-request-id header everywhere.
• Forward the incoming header through requestContext middleware instead of always minting a fresh ULID.
• Log under the unified requestId key so Datadog filters work across services.

Test plan

• Unit tests for requestContext middleware
• Manual trace from web → api → billing worker
• Verify Datadog dashboard once deployed to staging

curl -H "x-request-id: test-123" https://api.atlas.local/v1/ping
# logs: { "requestId": "test-123", ... } in all three services

Closes #1402 Refs #1390