[decorators] @cache never evicts; @rate_limit is process-global and raises bare ValueError #42

Summary

In nitrostack/core/additional_decorators.py, @cache never evicts entries and has no size cap, and @rate_limit raises a bare ValueError and shares one counter across all callers.

Findings

• cache(): cache_store is a per-function dict. An expired entry is only overwritten when the same key is requested again, so high-cardinality arguments grow the dict without bound.
• rate_limit(): one calls list per decorated function, so the limit applies to all users and instances together. It raises ValueError("Rate limit exceeded ..."), which a client cannot tell apart from a validation error, and it gives no retry hint.
• Both are in-process only.

Proposal

1. cache(ttl, max_size=...) with LRU eviction, plus a cache_clear() hook for tests.
2. A typed RateLimitExceededError carrying retry_after_seconds.
3. An optional key= callable on rate_limit for per-user limits.
4. Document that both are per-process.