New option: --per-file-ignores

[un-pogaz]

Close#3673

This implementation use nargs=2 to pass the file glob and the words list with a pair of arguments, rather than try to extract the data from a single complexe string. Also use action="append" so that several --per-file-ignores can be provided in one comand. This need to be a bit tricky to rebuild the resulting dict throught the config file.
Done a larger edit than targeted because of various Ruff warning about too many statements and complexity.

[Kristinita]

Type: Question ❓

@un-pogaz, have you considered implementing the option --per-context-ignore instead of --per-file-ignore? In this case, ignoring could be more accurate, see the issue #3673 for details.

Thanks.

[un-pogaz]

@Kristinita I suggest you to take a look at the already existing feature --exclude-file: each line into the given file will be ignore by codespell.
Since codespell ignore only the exact lines provided into the exclude file, and since the probability that your code will contain exactly the same line several times in different files, you can exclude some error with even greater precision than --per-file-ignore.

--per-file-ignore is an intermediate solution that target when --exclude-file will be too tedious to use (to many lines) and too risky to completely ignore the word globally with --ignore-words-list.