* volume: an already-deleted EC needle is not a delete failure
Deleting a needle that is already gone is what the caller asked for, and the
non-EC paths have always said so: BatchDelete reports StatusNotModified when
DeleteVolumeNeedle finds nothing to do, and DeleteHandler answers 404 from its
ReadVolumeNeedle pre-check. The EC branches had no such case, so ErrorDeleted
fell through to a generic failure -- 500 from both, and DeleteHandler also
counted it in VolumeServerFileWriteFailures, inflating a failure metric on a
replayed or duplicated delete.
The filer already tolerates this by string-matching "already deleted" on the
result, which leaves an error message load-bearing; the status is now right at
the source instead.
Claude-Session: https://claude.ai/code/session_01P3pE6J2UPFp6G3ksfMV4s1
* volume: close the EC fixture's disk location
Close stops the location's disk-space goroutine and releases the mounted EC
volume's file handles, which otherwise live until the test binary exits.
Claude-Session: https://claude.ai/code/session_01P3pE6J2UPFp6G3ksfMV4s1
