Fix allocated_capacity tracking when rescheduling

Fix allocated_capacity tracking when rescheduling

This aims to fix capacity tracking in the manager in case of
rescheduling. Idea is to inject information if volume was rescheduled
into the exception passed up so manager can decide if incrementing
allocated_capacity is needed. Then in the manager finally block is used
to make sure that allocated_capacity is incremented in case of failure
that haven't triggered rescheduling. Also adding two unit tests checking
if tracking is correct.

Closes-Bug: 1408763

Change-Id: Icd6b04ac60c17cbda225d3be8bcc4435ea0cd7a8