mem_scrub: propagate scrub wait failures

[TimZaman]

When the scrub work list is full, scrubSubmitPages() waits for older scrub work to complete before submitting more pages.

If _scrubWaitAndSave() times out, _scrubCheckAndSubmit() currently returns 0. The outer loop treats that as progress even though scrubListSize and free entries do not change, so it retries forever and repeatedly (trillions of times; 3k/s...) prints:

_scrubWaitAndSave: Timed out when waiting for scrub jobs to finish
NV_ERR_TIMEOUT ... @ mem_scrub.c:741

Before

_scrubWaitAndSave() returns NV_ERR_TIMEOUT
→ _scrubCheckAndSubmit() returns 0
→ scrubSubmitPages() subtracts 0, advances curPagesSaved, and retries
→ loop condition never changes
→ timeout/log repeats indefinitely

This PR

_scrubCheckAndSubmit() returns the original error plus actual submitted/saved counts
→ scrubSubmitPages() applies only real progress
→ timeout/error exits immediately instead of retrying
→ completed scrub work is still returned to PMA

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}