Motivation
To justify enabling the block cache in production we need to quantify the benefit it delivers. Today the blockCache ServerStatus section (pali_block_cache.cpp) reports only requests.total/cached/noncached and cache size/count — enough for a raw hit ratio, but not enough to show the IO/latency avoided by serving a page from cache instead of the page server.

Approach

• Add counters for bytes/pages served from the cache (hits) versus fetched from the page server (misses) in BlockCache::tryGet() / put() (pali_block_cache.cpp).
• Add an estimate of page-server round-trips and IO/latency avoided on hits (the disagg.pali.handle.GetCache AutoStat timings can inform this).
• Surface the new counters in the blockCache ServerStatus section so they flow into FTDC and can drive the Grafana panels.

Definition of Done

• New perf stat(s) (bytes served from cache, page-server fetches avoided, estimated IO/latency saved) appear under db.serverStatus().blockCache and in FTDC.
• A unit test in pali_test.cpp exercises hits and misses and confirms the counters reflect them.