Add noqa skips in otherwise empty lines to the next non-empty line

[StopMotionCuber]

Fixes: #3935

So what it does:

• Some magic in traverse_yaml to flatten lists within there (these are required for a ruamel CommentMap, please don't ask any further, afaik these are entirely undocumented)
• After collecting all #noqa and inserting them into the lintable, postprocessing is done on the lintable.
  • For each line where a skip is found, it is matched against a regex whether # noqa is the start of the line, minus some whitespace/indentation.
  • If there is only the # noqa on that line, the skip is added to the next non-empty line.

This is something required for me in order to suppress errors on yaml multiline strings, as reported by the issue referenced.

In general I think the solution is quite hacky, so this can also be seen as a proof of concept. The whole flattening thing and how ruamel works with comments is quite weird. As in the context of _get_rule_skips_from_yaml only a map of {lineno: rule_to_skip} is what we're interested in, it might be easier to just go over the content of a file, check each line for _noqa_comment_re match. The whole tree structure of CommentMap is annoying to traverse, a simple iterator over all comments would do it (don't know if that's in pyyaml?) and would be more accurate for comments not attached to objects in a line.

[ssbarnea]

This will have to wait until we finish the changes started in #4564 - adding support for ansible 2.19 with data tagging. There were a really big number of changes related to this and they also affect this area.

[StopMotionCuber]

That makes sense. I already noticed that PR. I didn't see any commit to skip_utils.py so I thought it was fine to go ahead, and I couldn't read the JIRA Ticket to get more information 😕

I'd happily contribute as soon as refactoring is done, just ping me again

[StopMotionCuber]

@ssbarnea I guess the data tagging changes are integrated now, any chance you could take a look again?