Fix race conditions in ObjectStorage where the s.index map was accessed without proper mutex protection in multiple locations.

This PR extracts and focuses on just the index map race fixes from PR #2094. The other fixes in that PR have been superseded by:

• PR plumbing: format/packfile, use ReadAt in FSObject.Reader() for concurrency safety. #1998 (FSObject.Reader concurrency)
• PR storage: dotgit, fix hasIncomingObjects race #2131 (DotGit.hasIncomingObjects)

Changes

storage/filesystem/object.go:

• Document muI and muP mutex fields to clarify what they protect
• Reindex(): Add mutex protection around s.index = nil
• PackfileWriter(): Add mutex protection in Notify callback when writing to s.index[h]
• EncodedObject(): Add read lock when checking if s.index != nil
• HashesWithPrefix(): Copy index values to slice while holding lock before iteration
• buildPackfileIters(): Add read lock when accessing s.index[h] in closure

storage/filesystem/object_race_test.go (new):

• Add TestConcurrentIndexAccess to reproduce the index map race conditions
• Test simulates concurrent readers (HashesWithPrefix, EncodedObject) and writer (Reindex)
• Run with go test -race -run TestConcurrentIndexAccess ./storage/filesystem/

Note on Test Output

The test may still report race conditions from external dependencies (e.g., memfs package). This is expected - those are separate issues outside the scope of this PR. The s.index races that this PR fixes are now resolved.

Testing

# Run the race test
go test -race -run TestConcurrentIndexAccess ./storage/filesystem/

# Run all filesystem tests with race detector
go test -race ./storage/filesystem/... -timeout 5m

Supersedes #2094