fsal: enforce folder quota on write/remove/truncate/rename

  Wires the QuotaManager into the mutating FSAL operations that can
  change tracked byte usage. Writes check the net new bytes up front
  and adjust the counter by the real delta after the write completes,
  so overwrites don't double-count. Remove captures the file size
  before unlink and releases it afterwards. setattr_size releases
  space when truncating down; extend is left alone since sparse files
  don't consume real bytes and the subsequent write() will be caught.
  Rename is a no-op within a single quota directory; across quota
  directories it checks the target's budget up front and transfers
  the usage atomically with the filesystem rename.

  NFS handlers for WRITE, SETATTR and RENAME now map the FSAL's
  "Quota exceeded" error to NFS3ERR_DQUOT (69), letting Linux clients
  surface the condition as EDQUOT instead of a generic ENOSPC.

  Also adds internal helpers quota_target / quota_dir_of on
  LocalFilesystem and a free-standing path_size that recursively sums
  file bytes under a directory for the cross-quota rename case.

Signed-off-by: Vicente Cheng <vicente.cheng@suse.com>

Author: Vicente-Cheng