Implement LRU status-list cache with in-memory and Redis backends

## Objective

Provide a pluggable LRU cache for recently used status-list records, with the backend chosen at compile time via Cargo features (`memory` or `redis`).

## Context

The server currently keeps status-list records in an in-process `moka` cache. For HA / multi-replica deployments, we need a cache abstraction that can be backed either by an in-process LRU cache or by Redis, selected when the binary is built.

## Deliverables

- [ ] Define a `Cache` trait exposing `get`, `put`, and `invalidate(list_id)` under a new domain ports module.
- [ ] `cache-memory` feature: LRU-capable in-memory implementation using `moka`.
- [ ] `cache-redis` feature: Redis-backed implementation with configurable TTL and invalidation.
- [ ] Ensure only one backend is active at a time; fail compilation if both or default to memory.
- [ ] Replace the current direct `moka` usage in `get_status_list.rs` and `update_status.rs` with the trait.
- [ ] Preserve the existing cache-invalidation semantics on `PATCH` updates.
- [ ] Add Prometheus-style hit/miss metrics for the cache.
- [ ] Add unit/integration tests for both backends.
- [ ] Update `docker-compose.yml`, `.env.template`, and Helm values with the new cache options.