Add structured log for each request that exceeds the configured response time threshold

[Chucks1093]

Summary

Slow requests are currently invisible until they appear in error logs or user reports. A warn-level log for any request exceeding a configurable threshold (default 2000ms) would give operators a proactive signal of performance degradation without requiring external APM tooling.

Scope

• Measure total request duration in the response middleware
• Emit a warn-level log when duration exceeds the threshold with fields: method, path, status_code, duration_ms, wallet (if authenticated), slow_threshold_ms
• Threshold configurable via environment variable SLOW_REQUEST_THRESHOLD_MS (default 2000)
• Log emitted after the response is sent, not before
• Do not log the request body or response body

Acceptance Criteria

• Warn log emitted for requests exceeding the threshold with all six fields
• Threshold configurable via environment variable
• Log emitted after response is sent
• Request and response bodies absent from log
• Fast requests produce no slow-request log entry

ETA: 12 hours

---

Coordinate on Telegram