Preserve updated Q after cache eviction

[Frankie-Xu]

MemoryService keeps both a fast _q_cache and a _mem_cache of memory objects. After a successful Q update, FIFO eviction can remove an ID from _q_cache; the next retrieval then falls back to that memory object's stale metadata.q_value and silently loses the learned value.

Synchronize the cached memory metadata after a successful non-None single or batch update. This keeps the existing in-process fallback consistent after Q-cache eviction without changing the Q-learning formula, FIFO policy, retrieval scoring, or cross-process behavior. Storage failures and unsuccessful updates do not mutate either cache.

Validation:

• On main c1b322c, the new single and batch eviction regressions fail (-0.5/0.5 learned values fall back to 0.0); failure and snapshot tests pass.
• With this patch, 4 unittest cases pass; the same 4 pass under pytest and from a clean git archive export.
• Coverage includes single and batch updates, FIFO eviction, negative and zero Q values, persistence failure without cache pollution, and q-cache snapshot round-trip.
• Tests execute the real MemoryService retrieval/update and QValueUpdater paths with only MemOS import types, in-memory persistence, and fixed embeddings substituted.
• compileall and git diff --check pass.

No live MemOS integration, cross-process consistency, concurrency redesign, or benchmark run is claimed. This is independent of PR #14's finite-value validation and PR #16's single-update Q-cache publication.