We already canceled running pipelines before. Gitlab, however, might
have already canceled previous pipelines automatically, but without
canceling their child pipelines.
Extend the mechanism now to cancel all pipelines up to and including
"running" as well as "canceled" states.

The list() call on Gitlab classes uses pagination per default. If the
number of items exceeds a certain limit, we only get a part of the
results. We can pass "all=True" to the calls in order to disable
pagination and get all results.
If we're looping over the items, though, the recommended way of getting
them is to use a generator object, which works calling list() with
"as_list=False". See documentation for reference:

https://python-gitlab.readthedocs.io/en/v2.10.1/api-usage.html#pagination

Note: the "as_list" argument is deprecated in newer versions of the
python-gitlab library, in favor of the newer "iterator=True":
https://python-gitlab.readthedocs.io/en/v3.6.0/changelog.html#feature
We are using v2.10.1 of the library, though, so this commit also updates
calls using the new way, which is not working with our version.

The old printf-style string formatting syntax using '%' is not
recommended anymore. See for reference:
https://docs.python.org/3/library/stdtypes.html#printf-style-string-formatting
https://docs.sourcery.ai/Reference/Python/Default-Rules/replace-interpolation-with-fstring/

Python 3.6 introduced f-strings as a better alternative:
https://docs.python.org/3/reference/lexical_analysis.html#f-strings

Consequently switch to this new syntax for all strings.

Everytime we push a new commit to a branch, a new pipeline is created.
Often the pipeline for the previous commit has not finished yet, and
GitLab does not automatically cancel it. It consumes valuable build
time, even if we don't need the results anymore.
Adding a MR pipeline job to search for previous pipelines on the same
branch and cancel them explicitly.