@hackage shikumi-cache0.1.2.3

The caching subsystem for shikumi: a content-addressed cache key (a BLAKE3 digest over a canonical request serialization — the contract EP-7's replay reproduces), a provider-neutral Cache effect, an in-memory backend, and the cachedLLM memoizing interpreter that wraps EP-1's LLM effect so identical requests contact the provider once.