feat: implement adaptive rate limiting and throttling system

[sonoflawal]

closes #176

Summary

Implements an adaptive rate limiting system that dynamically adjusts rate
limits based on real-time platform health signals, consumer behaviour, and
detected anomalies — complementing the existing static limits.

What's changed

• Signal collection — samples CPU utilisation, DB pool utilisation, Redis
  memory pressure, request queue depth, error rate, and p99 response time on a
  configurable interval with rolling averages to smooth transient spikes

• Adaptation modes — Normal → Elevated → Critical → Emergency, driven by
  configurable per-signal thresholds with hysteresis to prevent oscillation

• Consumer priority tiers — High (partners, microservices) are protected
  from tightening in Elevated/Critical; Standard and Low receive progressively
  aggressive multipliers; all tiers get a reduced but non-zero limit in Emergency

• Per-consumer trend tracking — detects accelerating request rates and
  applies an additional throttle multiplier during Elevated/Critical mode

• Graceful degradation — non-essential endpoints return 503 with
  Retry-After in Critical/Emergency; essential financial endpoints are queued
  up to a configurable depth with FIFO shedding when full

• Admin control — GET/POST/DELETE /api/admin/adaptive-rate-limit/...
  endpoints for status inspection and manual mode override

• Redis sync — effective multipliers are pushed to Redis on every cycle so
  all application instances apply consistent limits simultaneously

• Observability — Prometheus gauges for mode, all signal values, rolling
  averages, and effective multipliers per tier; counters for mode transitions,
  per-consumer throttle applications, and request shedding events; structured
  log events with full signal context on every Critical/Emergency transition

Files

Path	Purpose
src/adaptive_rate_limit/	New module (12 files)
migrations/20260403000000_adaptive_rate_limiting.sql	Signal snapshots + mode transitions tables
src/metrics/mod.rs	Registers adaptive rl metrics in global registry
src/main.rs	Wires up worker, middleware, and admin routes
src/lib.rs	Exports new module
tests/adaptive_rate_limit_integration.rs	Integration tests

Testing

• 20 unit tests covering rolling averages, threshold evaluation, hysteresis
  enforcement, multiplier application per tier, accelerating request detection,
  request shedding logic, and endpoint classification
• Integration tests covering mode transitions, manual override, multiplier tier
  ordering, hysteresis, and worker lifecycle

[drips-wave]

@sonoflawal Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits