feat(backend): introduce bounded lru_cache and observability to query

[SaitejaKommi]

Summary

This PR modernizes query-builder caching in search.py by replacing manually maintained unbounded dictionaries with bounded native lru_cache strategies.

It introduces automatic config-aware cache invalidation and exposes runtime cache metrics for observability, improving memory safety, maintainability, and operational tuning.

---

Related Issue

Fixes #401

---

What Changed

Refactored Cache Engine

Removed manually managed in-memory cache dictionaries:

• _ES_QUERY_BUILDERS
• _RESULT_PROCESSORS

Introduced Bounded Caching

Wrapped query-builder / processor factories using:

@functools.lru_cache(maxsize=32)

Benefits include:

• Prevents unbounded memory growth
• Uses proven Python-native cache semantics
• Simplifies custom cache management logic

Config-Aware Invalidation

Bound cache keys to a configuration identity/version parameter using:

id(get_config())

When the active config object reloads, the new identity naturally causes cache misses and rebuilds fresh entries.

Added Observability Hooks

Exposed runtime cache metrics via:

.cache_info()._asdict()

Surfaced values include:

• hits
• misses
• maxsize
• currsize

These metrics are now available through health diagnostics, enabling live tuning and monitoring.

Improved Debug Visibility

Added debug logging around Elasticsearch translation / cache evaluation paths.

---

Affected Files

• search.py
• health.py
• Related cache behavior tests

---

Testing

Verification gateway executed successfully:

• Backend unit tests ✅
• Backend integration tests ✅
• Frontend build/tests (dev) ✅
• Frontend build/tests (prod) ✅

---

Notes

This is a backend performance and observability enhancement with no intended API behavior changes.