Refine string case conversion helpers

[arnaudgallou]

Fixes #592

Let me know if you want me to add short comments explaining each regex pattern.

I'm not quite sure how you'd like to handle acronyms in str_to_camel(). I'm more inclined to treat acronyms like any other words (e.g. userId; current behavior), but I know some people prefer preserving uppercase in two-letter acronyms (e.g. userID).

Some tests might be redundant.

[arnaudgallou]

I read a bit more on the differences between POSIX pre-built classes and unicode regex. The main difference is that POSIX classes are locale-dependent when unicode regex are not (see https://www.regular-expressions.info/posixbrackets.html).

I wasn't able to test it but POSIX classes may be able to handle the ij digraph properly if the locale is set to Dutch.

[arnaudgallou]

Also, how would you like to handle superscript and subscript letters and numbers? Should we drop them, or normalize them using stringi::stri_trans_nfkd()? Normalizing the input string could be useful for handling certain units like m², which also raises the question of how to treat digits.

Currently, digits are always separated from letters. Another option would be to keep digits attached when they directly follow a letter, but use the dash or underscore separator otherwise. The downside of this approach is that it makes the conversion from snake/kebab case to camel case non-reversible.

[hadley]

Thanks! Lets keep this simple for now, and not worry about special characters (subscripts etc).

I could also imagine options that ensure that the output is ASCII (since you might want that for programming variables) or valid R variable names, but I think extensions can wait until we see some use of this in the wild.