[#099] stale-while-revalidate cache dedupe

[Menjay7]

Summary

This PR introduces a stale-while-revalidate (SWR) caching strategy with request deduplication, reducing redundant backend calls and improving response latency under concurrent load. It ensures that identical in-flight requests share a single computation while serving stale cached data immediately.

Changes Made
Implemented stale-while-revalidate caching layer.
Added in-flight request deduplication (single-flight protection).
Cached responses are served immediately while background refresh updates stale data.
Introduced TTL-based cache expiration with soft and hard expiry separation.
Added request key normalization to ensure consistent cache hits.
Integrated cache invalidation hooks for critical mutations.
Added metrics tracking for cache hits, misses, and deduplicated requests.
Problem

Without deduplication and SWR, multiple identical requests:

trigger duplicate expensive backend computations
increase database/API load
degrade performance during traffic spikes
Solution

This PR introduces a combined strategy:

Serve stale cache instantly for fast responses
Revalidate in background without blocking requests
Deduplicate concurrent requests using an in-memory promise registry

This ensures:

Only one upstream request per key at a time
Other requests reuse the same pending result
Cached data remains responsive even during refresh
Key Features

1. Stale-While-Revalidate
   Returns cached response immediately (even if stale)
   Triggers background refresh asynchronously
2. Request Deduplication
   Prevents duplicate in-flight executions
   Uses a shared promise map keyed by request hash
3. Cache Strategy
   Soft TTL: data can still be served while stale
   Hard TTL: forces recomputation
4. Observability
   Cache hit/miss tracking
   Deduplication counters
   Revalidation timing metrics
   Testing
   Unit Tests
   Cache hit returns stored value immediately
   Stale cache triggers background refresh
   Duplicate concurrent requests return same promise
   Hard TTL forces recomputation
   Cache key normalization correctness
   Integration Tests
   Multiple parallel requests deduplicated correctly
   Backend called only once per unique key
   Cache invalidation after mutations
   SWR behavior under load
   Impact
   Reduced backend load under concurrent traffic
   Faster average response times
   Improved system stability during spikes
   Lower database/API cost
   Better user experience with near-instant responses
   Type of Change
   Performance Optimization
   Infrastructure Enhancement
   Caching Layer Improvement
   Reliability Improvement...Closed [#099] stale-while-revalidate cache dedupe #190

[vercel]

@Menjay7 is attempting to deploy a commit to the ezedikeevan's projects Team on Vercel.

A member of the Team first needs to authorize it.

[Menjay7]

Both PR and conflict are in check sir

[Menjay7]

Done sir..@ezedike evan