In memory cache

[gloskull]

Motivation

• Provide a deterministic, dependency-free in-memory cache for critical paths with configurable TTL and capacity to reduce upstream load and meet latency budgets.
• Allow off-chain services to swap in a Redis-backed adapter without changing call sites by carrying Redis deployment settings alongside the in-process config.
• Expose basic observability (hits/misses/evictions/expirations) and a rollout/runbook to support blue-green/canary deployments and monitoring.

Description

• Add a new cache module src/db/cache.rs implementing CacheConfig, RedisCacheConfig, CacheMetrics, and a generic TtlCache<K,V> with per-entry TTL overrides, expiry on access, bounded capacity eviction, and metrics.
• Export the new module via pub mod cache; in src/db/mod.rs and add a brief crate-level comment referencing the cache in src/lib.rs.
• Add documentation CACHE_LAYER_ARCHITECTURE.md describing TTL policy, Redis deployment guidance, monitoring/alerting, rollout guidance, and runbook steps.
• Include comprehensive unit tests in src/db/cache.rs covering TTL expiry, per-entry TTL overrides, capacity eviction, metrics, and Redis config defaults.

Testing

• Ran cargo fmt and cargo test cache --lib to exercise the new cache unit tests and related library tests.
• All executed unit tests passed (17 tests ran, 17 passed, 0 failed).
• Verified formatting with cargo fmt before final tests.

Closes #71