Change `skip_broken` and `fail_on_warnings` options

[neomilium]

From my point of view, we should change our approach about two options skip_broken and fail_on_warnings:

• skip_broken is not relevant option as, for example, an "update" that does not contain modification is consider as broken
• skip_broken sounds me to be the default so turning option from skip_broken to fail_fast makes me sense to me
• fail_on_warnings (ie. exit code sets to 1 or higher) could be an option if we choose to introduction a distinguish between real error (e.g. unable to process a template) and a warning (e.g. no files are updated after an update run)
• finally, we should always fail earlier (with an exit code sets to 1 or higher) if an error occurred, so fail_fast will be relevant in case of warnings

[neomilium]

What are the users opinion?

[smortex]

msync raising an error when a repository is up-to-date with the expectations if highly irritating… Having to pass skip_broken makes me think that I might fail to spot real problems for modules managed after the up-to-date module.

I consider this to be a bug indeed, and fixing this bug basically makes msync behave as make: by default stop on the first error, but by passing extra arguments (--keep-goind / -k)allow to continue. I quite like this behavior, but have no strong opinion about toggling it. In this case, adding a summary of the failures at the end (like rspec does), might be a good thing though.