feat(caching): improve cache warming with limits, batching, metrics, …

[abdulazizishaq212-prog]

Addresses five weaknesses in the cache warming pipeline: unbounded DB reads, unweighted entry selection, burst Redis
writes, no observability, and warming that competed with the boot sequence.

Changes

Configurable entry ceiling (CACHE_WARM_MAX_ENTRIES)

Every DB read in the warming pipeline now has an explicit take / limit derived from CACHE_WARMING.MAX_ENTRIES (default
500, overridable via CACHE_WARM_MAX_ENTRIES). Previously, warmCoursesList issued an unbounded find against the courses
table.

Real-signal prioritisation

• Courses list — ranked by enrollment count descending (same signal used by popular courses), with a createdAt DESC
  recency fallback when the enrollment table is empty.
• User profiles — ranked by lastLoginAt DESC so the most recently active users fill the warmed set; falls back to
  updatedAt DESC for fresh installs or seed data.

Batched processing with inter-batch delay

warmUserProfiles now processes the working set in slices of CACHE_WARMING.BATCH_SIZE (default 50, env:
CACHE_WARM_BATCH_SIZE), with a CACHE_WARMING.BATCH_DELAY_MS pause (default 50 ms, env: CACHE_WARM_BATCH_DELAY_MS)
between batches. This bounds peak Redis write throughput and heap allocation per warming cycle.

Prometheus metrics

Two new instruments added to MetricsCollectionService:

• cache_warming_entries_total — Counter, labelled by target — total keys written per warming run.
• cache_warming_duration_seconds — Histogram, labelled by target — wall-clock duration per run.

Every warm* method calls metrics.recordCacheWarming(target, keysWarmed, durationMs) after completing.

Deferred startup warming

CacheWarmingScheduler now implements OnApplicationBootstrap instead of OnModuleInit. NestJS fires onApplicationBootstrap
after the HTTP server is listening, so warming no longer competes with the tail end of the boot sequence. An additional
CACHE_WARMING.STARTUP_DELAY_MS grace period (default 5 s, env: CACHE_WARM_STARTUP_DELAY_MS) further defers the first
warmAll() call.

Environment variables (all optional)

┌─────────────────────────────┬─────────┬──────────────────────────────────────────┐
│ Variable │ Default │ Description │
├─────────────────────────────┼─────────┼──────────────────────────────────────────┤
│ CACHE_WARM_MAX_ENTRIES │ 500 │ Max rows fetched per warming target │
├─────────────────────────────┼─────────┼──────────────────────────────────────────┤
│ CACHE_WARM_BATCH_SIZE │ 50 │ User profiles processed per batch │
├─────────────────────────────┼─────────┼──────────────────────────────────────────┤
│ CACHE_WARM_BATCH_DELAY_MS │ 50 │ Delay between profile batches (ms) │
├─────────────────────────────┼─────────┼──────────────────────────────────────────┤
│ CACHE_WARM_STARTUP_DELAY_MS │ 5000 │ Grace period before startup warm-up (ms) │
└─────────────────────────────┴─────────┴──────────────────────────────────────────┘

Testing

• All 15 unit tests pass (pnpm jest cache-warming).
• New tests cover: enrollment-count prioritisation, recency fallback, BATCH_SIZE enforcement (via fake timers), deferred
  startup delay assertion.
• Build passes (pnpm build).

Files changed

• src/caching/caching.constants.ts
• src/caching/cache-warming.service.ts
• src/caching/cache-warming.scheduler.ts
• src/monitoring/metrics/metrics-collection.service.ts
• src/caching/cache-warming.service.spec.ts
• src/caching/cache-warming.scheduler.spec.ts

closes #1053

[drips-wave]

@abdulazizishaq212-prog 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

[RUKAYAT-CODER]

Well done on the job done so far!
Kindly fix workflow to pass.