Our Perl images are now all based on Debian Buster so there is no longer
a need to try and install Jessie packages.

Seeing as they were to run even in the event of any of the local
test jobs having failed, no real point in waiting until after the local
jobs have completed.

What I have done here is take .gitlab-ci.yml and
MultiTestDB.conf.gitlabci.mysql from ensembl (my own development version
in case of the former), modify the before_script section of the former
to account for differences in dependencies between the two projects,
create a dependent-build trigger attempting to launch one on
ensembl-gh-mirror/ensembl (i.e.  already on EBI Gitlab), and fix the
declaration of same because it has turned out variable expansion
in trigger declarations works only in the branch field, not in "project".

Note that all of this results in a somewhat different build
matrix than on Travis but unlike there, on GitLab we have got
consistency between ensembl and ensembl-test.

patch_test_databases.pl: add an option for restricting the patching to specific DB types

Added the biotype table to the list of "global" tables

Note that said option takes precedence over the hard-coded list of types
to exclude so it _is_ possible to, should one be inclined to do so,
attempt the patching of 'hive' and 'web' databases.

Issue: ENSCORESW-3259

It is needed in order to get the bioytpe group of a gene / transcript

Bump test-database schema versions to 99/100

This updates the core test DBs from 90 to 100 and the compara one from
90 to 99. Moreover, we now actually have SQLite test DBs instead of
empty directories.

Fixed bug when core dbs don't have contigs

Removed assumption that the 3 levels of assembly are chromosome,
supercontig and contig, as there are many other species with different
coord_system sequence level (e.g. primary assembly, scaffold).

A similar error might be raised in the future regarding supercontigs,
and since 'toplevel' and 'seqlevel' are linked in the assembly table,
that part of the code has also been removed.

Travis: explicitly use Ubuntu Trusty for builds

Matthieu Muffato's experiments with ensembl-hive have shown that in
spite of https://docs.travis-ci.com/user/languages/perl/ claiming
Travis supporting perl-5.8 and newer, the oldest version supported under
the Xenial environment is 5.22. This has since been confirmed by a post
on Travis community forums:
https://travis-ci.community/t/failure-with-perl-5-16-5-18-5-20/2458 .

We are presently not ready to drop support for Perl versions older than
what Travis makes available under Xenial, therefore let us for the time
being make sure we continue to use Trusty for now.

Add a README with the Travis badge

Travis: Add mysql to the list of services [master]

Ubuntu Xenial, which is now being rolled out as the default Linux distro
for Travis builds following the end of life of Ubuntu Trusty, does not
start services such as MySQL unless told to. We use the default distro
in our builds so we are likely to be affected.

[ENSCORESW-3077] Travis: Unify configuration of Slack notifications

Make sure Travis installs IO::Scalar before building Net::FTPServer

See the comment in .travis.yml for details.

Move dump_test_schema.pl dependencies from a README to a cpanfile feature

It is mostly redundant now that dependencies of dump_test_schema.pl are
in the cplanfile, moreover it was out of date (MooseX::App:Simple has
since become a dependency of ensembl-test as a whole).

The aforementioned script, which is used in the course of patching test
databases, has got two additional dependencies (which in turn depend on
quite a few other new modules), which until now were only mentioned in
a dedicated README file in scripts/ . Put them in the cpanfile where
they belong, wrapped in a dedicated feature so that we do not inflict
them on people who do not want to ever patch test databases.

Custom prefix for MultiTestDB database names

Until now, MultiTestDB would prefix the names of test databases it
creates with the local user name as determined by looking at the
environment variables: LOGNAME and USER, in the order as written here.
There are the following problems with this approach:
 - it leaks local user names to the database system and by extension to
   everyone who can access them,
 - in the event of said local user names being different from farm ones
   it makes it more difficult to identify to whom a specific database
   belongs.
To address these issues the present commit introduces a new environment
variable called ENSEMBL_TESTDB_PREFIX which when defined takes
precedence over LOGNAME. Now anyone running tests on their local machine
can set that variable to their farm user name.

Yearly copyright update

allow new licence format without copyright year

Modify script to create collection databases for testing.

Enable the script to create collection databases for testing. The code already iterates over multiple species in a single json config file, so just need to note when the first one is done, then avoid doing db creation and 'global' table population on subsequent iterations.

[ENSCORESW-2622] DBIx warning in patch test databases