Rather than call 'fnic_flush_tx()' from interrupt context we should be
moving it onto a work queue to avoid any locking issues.

Fixes: 1a197555 ("scsi: fcoe: Fix potential deadlock on &fip->ctlr_lock")
Co-developed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Lee Duncan <lduncan@suse.com>
Link: https://lore.kernel.org/r/ce5ffa5d0ff82c2b2e283b3b4bff23291d49b05c.1707500786.git.lduncan@suse.comSigned-off-by: Martin K. Petersen <martin.petersen@oracle.com>