The getNextSucceededRetrieveRequestForRepackBatch is implemented and integrated in the expandRepackRequest test

Started to implement the method to get the RetrieveRequest from the RetrieveQueueToReportToSuccessForRepack queue

Conflicts:
	xrootd-ssi-protobuf-interface

Refactored OStoreDB::RetrieveMount::batchSucceedRetrieveForRepack() method

Make was interpreting the variable instead of the underlying shell

-> Moved code related to payload manipulation in the cta::objectstore::RetrieveRequest class

Changed the operator "<" of Scheduler::PotentialMount so that the sorting of mounts is in alphabetical order

This was a rare occurence:
Tue Jan 29 01:12:51 2019 pid=531 tid=951 op=In BackendRados::lockBackoff(): radosCtx.lock_exclusive() obj=Frontend-ctafrontend-531-20190128-18:02:44-0 duration=51.5958

(One out of 5 operations taking more than 1s in the whole day.

The race is in EOS between MGM invalidating disk copies in the FST after garbage collection and
FST reporting new disk copy after retrieve.

Conflicts:
	scheduler/OStoreDB/OStoreDB.cpp

Upgraded to EOS version 4.4.20-20190125160957gitfec7ad8 in order to get the better performance version of the MGM LRU tape aware garbage collector

Changed the status of succeeded retrieve requests as RJS_Succeeded and inserted them to the RetrieveQueueToReportToRepackForSuccess, unit tested but memory leak

This reverts commit 5e55c15f.

Conflicts:
	mediachanger/acs/AcsTest.cpp
	mediachanger/acs/CMakeLists.txt
	tests/CMakeLists.txt

It allows running the unit tests in a build environement with:
make -C CTA-build/tests/ -f parallelTestsMakefile.local

This was based on the assumption that queue would be trimmed fast, leading to a tight loop. We now go one round on existing queues.

Attempting to lock a non-existing object leads to the creation of a zero-sized one. When this happens, the
locking code deletes the freshly created object immediately. Nevertheless, this prevented the creation of a
new object with this name. The creation code now retries to overcome this race.
In addition, reads finding a zero-sized object also consider it as absent.

implementation of "cta-admin archivefile ls -all".  A
missing order by clause meant that non deterministic
listings were given.

This commit adds the missing "order by" clause.

Status of repack retrieve request changed to RJS_Succeeded, started to insert the repack retrieve request into RetrieveQueueToReportToRepackForSuccess

all containers.
Many failures were due to slow DNS name resolution by kube-dns.
To solve this I am now scanning all registered services in the namespace
and adding entries to /etc/hosts on all the `Running` containers in the
namespace. This solution is valid for both systemd and non-systemd
containers.