Should rome check --apply{,-suggested} report unfixed errors and exit 1 if any?

[jeysal]

Coming from #3767 (comment), where we discovered that there is not currently an easy replacement for eslint --fix x.js, because rome check --apply x.js does not print unfixed errors and exit 1 if there are any.

I'm suggesting that rome check --apply, like rome check without --apply, print the remaining errors after applying all safe fixes, and exit with code 1 if there are any. Likewise, rome check --apply-suggested should print remaining errors after applying all safe and unsafe fixes, and exit with code 1 if there are any.

This not only aligns with ESLint, but also makes more sense in terms of CLI design IMO. A subcommand should always do in principle the same thing, and flags should only alter its behavior but not make it do something else entirely.
As is, rome check --apply is actually more of a separate command (like a rome apply or rome fix) because it has no overlap in FS outputs or in Console output with rome check without --apply (one makes no file changes and outputs a list of errors, one makes possible file changes and outputs a summary of them).
The suggested rome check would always output a list of errors, and adding the --apply flag would (logically speaking) alter it to first try to fix errors before then proceeding with the normal rome check behavior.