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

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.

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.

Following on from GitLab issue cta/CTA#371, replaced INTEGER with NUMERIC(20, 0) within oracle_catalogue_schema_header.sql

Following on from GitLab issue cta/CTA#371, replaced INTEGER with NUMERIC(20, 0) within mysql_catalogue_schema_header.sql

Changed common_catalogue_schema.sql for cta/CTA#371 'Change INTEGER columns of CTA catalogue schema to NUMERIC(20,0)'

While fixing #360

moving core dumps to /var/log/tmp/%h-%t-%e-%p-%s.core so that they are available as artifacts for all pods AND in the log PV that has supposedly more space.

delete_instance.sh can write log to an alternate directory and not only /tmp as it may be too small for larger tests.

Adding systemd-resolved to all possible DNS intense pods to avoid Nameserver complaints for larger tests. client pod is now using systemd as well for this reason.

1 block per file and blocks >=256kB: more smaller blocks (min size). Do not mount a tape for less than 100 files.

Fixed cta/CTA#364 RdbmsArchiveFileItorImpl constructor should NOT return database connection to pool