Skipped AFS writes via cleared inode marking

Writeback state handling in affected Linux Kernel versions beginning with 6.14 leaves skipped AFS directory or link writes without a retry marker. When lock contention and WB_SYNC_NONE make netfs_write_single() or afs_single_writepages() skip a write, the VFS may clear the inode mark and the old code neither reports the skip nor re-marks the inode. The directory path is described as theoretical because afs_single_writepages() holds validate_lock, serializing access; no attacker-controlled trigger or concrete security consequence is published.