scsi: ufs: Avoid dead lock during error recovery

The UFS error handler work may be entered from the clock scaling context
which may have acquired the lock used to synchronize between clock scaling
and shutdown. Part of the error handler work is to send management
requests during which the same synchronization lock is acquired again.
This condition leads to a dead lock. To avoid this dead lock,
skip acquiring the synchronization lock when sending management
requests as part of the error handler.

Change-Id: I38179528868aaeb4a788a83ca8f9c458afea862a
Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>