error handling in sd suspend

Hi,

this patch series introduces proper error handling into sd suspend handling. This fixes several issue:

- errror codes are sanitized so that SCSI errors don't get into generic PM code
- the error returns from synchronizing the cache are evaluated and translated for the upper layer, this particularly helps when unplug races suspend eSATA is likely to do this, as it takes very long to handle an unplug.
- the error returns for stopping the unit are likewise sanitized

Regards
Oliver