Fix running of Ansible playbooks shipped with tmt

[falconizmi]

Plugins have to mangle playbook paths to appease a piece of code that
expected paths to always be relative and under the fmf root. This is not
true for playbooks shipped with plugins that are bundled with tmt, like
prepare/feature. Therefore changing the code to support plugins that
may never be under the fmf root.

Pull Request Checklist

• implement the feature
• write the documentation
• extend the test coverage
• update the specification
• adjust plugin docstring
• mention the version
• include a release note

[falconizmi]

This PR blocks #2985

[martinhoyer]

+1 to @happz's comments

[therazix]

It looks like this PR also fixes the #3032.

[happz]

It looks like this PR also fixes the #3032.

Hm, yes and no. Yes, I think this PR might fix #3032, because the plugin and tmt will now use a better way when speaking about paths of playbooks that are shipped with tmt. But also no, because the cause is not the one mentioned in #3032, i.e. a missing fmf root - that's actually a second issue. prepare/epel had to perform some black magic mangling of its playbook paths, which did involve fmf root, and this PR removes this magic, therefore fmf root is no longer involved, therefore prepare/feature will work. But not because tmt suddenly learned to run any playbook without having an fmf root - when it comes to user-provided playbooks, like those for prepare/ansible or finish/ansible, they will still fail outside of fmf root. We discussed both issues with @falconizmi and agreed on addressing them both, in separate PRs.

[thrix]

marked as must, users are hitting this in Testing Farm

[happz]

Assuming @falconizmi was not available on such late notice on Friday afternoon, I tried to address the comments above. I think there will be failed tests, I'll check them in the morning.

[happz]

And before the merge, $SUBJ & commit message should be updated - it's not fixing "outside fmf root" issue, it's fixing "feature can't run its own playbooks".

Done.