CI: conditionally run tests based on what changed

[timotheecour]

proposal

• add logic to detect whether we should run a test or not depending on PR content: if we can prove that a PR won't affect a test, skip it
• but still run all tests on a commit (as opposed to pull_request), as a safety net in case the above logic is wrong
• also provide a [ci force] flag to avoid skipping tests in a PR
• the logic should be reusable and written in nim so that all CI pipelines can reuse it

benefits

• faster CI turnaround times
• we could bring back openbsd testing in pull_requests if we show this brings down CI testing time on this platform to a suitable level (Excessive use of builds.sr.ht nim-lang/Nim#17479)
  refs builds.sr.ht: Split OpenBSD job causing resource usage concerns nim-lang/Nim#17107 (comment)

links

past efforts related to this: note that a suitable approach should use nim code instead of bash scripts

• [TODO] [CI] revive stale #6999 : Dont to run the tests if only the docs were updated; travis only for now by timotheecour · Pull Request #8552 · nim-lang/Nim
• CI: Don't to run the tests if only the docs were updated (WIP) by data-man · Pull Request #6999 · nim-lang/Nim
• Excessive use of builds.sr.ht · Issue #17479 · nim-lang/Nim
• refs fixes #17437 - crash where error reporting > 1 nim-lang/Nim#17547 (comment)

[WIP] fix #6999 (allow [nimDocOnly] in commit msg); fix #13477 (flaky nodejs install); log PR url (+other info) in runCI logs by timotheecour · Pull Request #13556 · nim-lang/Nim