Re-work Cache->Cursor handling in CacheDestroySlab()

Re-work Cache->Cursor handling in CacheDestroySlab()

Since the advent of lazy slab spilling in commit 7f8b622668fb ("Improve the
performance of the slab allocator"), if the cursor slab is being destroyed,
it no longer means it is the only slab in the last. Remove the ASSERTions
and simply set the new cursor to the current cursor's Flink. This will
either be the next slab which (because slabs are kept in decreasing order
of occupancy) will also be empty, or it will be the list anchor (which
indicates that all slabs in the list are full). Call CacheAudit() after
slab removal to re-verify these invariants.

Also clean up a typo in a commit comment in CacheAudit().

Signed-off-by: Owen Smith <owen.smith@cloud.com>
[Re-worked original patch]
Signed-off-by: Paul Durrant <pdurrant@amazon.com>