vifib-startup playbook (which contains only vifib-server role) should
run without any dependency on Internet since it runs at boot time on all
vifib servers and must not fail.

See merge request nexedi/slapos.package!179

This command should not block the playbook when a repository
is broken

CentOS has exception, so we keep installing with tasks

If for example a repository key is expired, this should not
block the whole processing, except when the package state is
"latest".

To achieve this, before installing package, we always try to
update apt cache in a separate step for which we tolerate
errors and in a second step we only install the package without
updating cache - unless the package state is latest.

The idea is when we want to install the latest package, it's a
problem if we can't update cache before, but when we just want
the package to be present, errors during apt update should are
not a problem.

Only debian was adjusted with this logic.

fixup for 203927d9

the filename was not OK (only used for Arch)

no more special mount directory for slapos stuff

update the slapos configuration

It seems some machines don't have the default 666 permission on
/dev/net/tun so we force it in the playbook.

Also add names to the tasks.

Since re6st is giving a /25 subnetwork to the machine and address 1 is
assigned to the host by re6st, it means we have only 125 addresses left
for tap interfaces.

See merge request nexedi/slapos.package!176

Cleanup unused and not working packages installations (socat)

See merge request nexedi/slapos.package!175

Fix random error "No space left on device".
For documentation, see
https://openbuildservice.org/help/manuals/obs-user-guide/cha.obs.build_job_constraints.html
For a case on OBS forum, see
https://forums.opensuse.org/t/obs-build-build-vm-line-427-echo-write-error-no-space-left-on-device/140717

Support MCA and fluent-bit plugin in separate packages.
Update command-line calls to the programs and files.

Update according to the .cfg file names in slapos repository.