reusePromptPrefix property

Reuses matching prompt prefixes from previous requests in the same native context to reduce prompt ingestion latency.

This optimization applies to native text-only generation. Exact full-prompt replays of an in-process cache are conservatively re-ingested. After loading a KV state file, native llama.cpp instead re-evaluates the final token on an exact match; scores and continuation can differ from full-prompt evaluation. Set this to false to explicitly evaluate the full prompt again, trading away prefix reuse.