Changelog for thread-utils-context-0.4.0.0

Changelog for thread-utils-context

Replace striped-IntMap internals with a flat open-addressed hash table backed by per-thread IORefs. Reads and writes on the hot path are now plain IORef operations, with zero CAS and zero contention.
Add CMM primops (stg_getCurrentThreadId, stg_probeThreadSlot) to eliminate ThreadId allocation and FFI overhead on the hot path.
New construction function: newThreadStorageMapWith for explicit capacity.
New getCurrentThreadId reads CurrentTSO.id directly via CMM.
New ref-based API for instrumentation hot loops: ensureRef, ensureRefFast, lookupRef, lookupRefFast, readRef, writeRef, modifyRef.
Remove containers dependency.
Requires cabal-version: 3.0 (for cmm-sources).
Backwards compatible: all previously exported symbols retain their original type signatures.