In DriveState::addDiskSpaceReservation and substractDiskSpaceReservation, use a reference to the mutable_disk_space_reservations item

Retry queueing in ArchiveQueueToReportToRepackForSuccess if switchElementsOwnership fails because of a rados::lockbackoff() problem

Added repack ls command at repack failure in repack system test

Corrected cta-objectstore-initialize program. The agent needed to initialize the objectstore was not locked so could not be deleted.
If this program fails for whatever reason, it will return EXIT_FAILURE and not EXIT_SUCCESS

Corrected auto garbage collection of ArchiveRequest ToTransferForRepack (getQueueType missed one case)

Increased the time to lock a Rados object from 300ms to 1 sec. This correction is due to CI repack error (caused by BackendRados::lockBackoff(): timeout)

Removed objectstore initialization agent after objectstore initialization (cta-objectstore-initialize.cpp)
Catched other exceptions in continuousintegration/orchestration/tests/client-ar-abortPrepare.cpp

Disk reservation where cleaned up in states when they should be kept.
Added disk reservation logs.

Avoided counting an error for files hitting end of tape.

Added activable logs to the RetrieveQueueStatistics cache for debugging

Added a check for the repack just add copies workflow that verify that two archive routes does not point to the same tapepool

Remove the repack buffer if the repack expansion indicates that no Retrieve is necessary
Corrected the XrootDirectory::exist() method

[REPACK] Added a failed to create archive request counter so that the Repack Request is complete or failed if Retrieve subrequest is failed (creation or execution)

Fixed various issues detected in the process.

The sleep time (15 minutes) is currently hardcoded, but implemented in a single place (OStoreDB::fetchMountInfo() ).

The process can now indicate the agent structure can be garbage collected, saving a wait for a timeout. This
is typically hapenning when archive jobs need requeueing after a session hits the end of the tape.

Started the strcutures needed for sleeping a retrieve queue.

Moved the space reservation information to the DriveStatus object store object instead of a new central registry.
The central registry would have been a single point of contention as was the DriveRegistry before being split into
DriveStates. As the problem is so close to the one of the drive status, we can actually reuse the drive status for this purpose.

The algorithm will also change as we move the responsibility of querying the free space from the disk systems into the OStoreDb
object instead of the Scheduler. This leads to a slightly worth layering of responsibilities, making the OStoreDb::RetrieveMount
object a client of the disk::DiskSystemFreeSpaceList object.

The current implementation will also query the free space from the disk systems on each pop, instead of doing so in a globally
cached fashion. With the new model, we could cache the free space per drive (if needed), but not globally. This is not expected
to be a real issue and free space is a global counter in the disk system, expected to be readily available.

Added mount policy to the Repack Request

Added timer for ContainerTraits<ArchiveQueue>::trimContainerIfNeeded() and ContainerTraits<ArchiveQueue>::getLockedAndFetchedNoCreate()

Added microseconds after the creation time of the object (Agent Reference)