Prototype a tree-sitter grammar for md + Rd

[hadley]

Would allow us to move away from the current approach of commonmark + regular expressions to protect Rd tags. Would eliminate a wide class of weird edge cases and might be faster.

[DavisVaughan]

I believe this would be beneficial for a number of reasons beyond use in roxygen2 itself
r-lib/tree-sitter-r#68

• When parsing an R file, we could inject the roxygen2 grammar into roxygen2 comment blocks. This could give us a way to add syntax highlighting to roxygen2 blocks

• When parsing the @examples inside a roxygen2 block, we could do a second inject of the R grammar. This may be able to give us syntax highlighting of the R code within @examples sections, but I'm not sure.

I would suggest that tree-sitter-roxygen2 should only be in charge of parsing the docs after all #' have been stripped out. In other words, tree-sitter-r is in charge of parsing the whole #' <content> line, and then tree-sitter-roxygen2 is in charge of parsing <content>

If you're wondering how multiple lines get combined into a "single" document to be reparsed by tree-sitter-roxygen2, I believe the injection.combined pattern can help with that
https://github.com/tree-sitter/tree-sitter/blob/8500e331ebfd49e66dd935b8a9c7a58aba68af37/docs/section-4-syntax-highlighting.md?plain=1#L370-L371

i.e. for

#' @param foo this
#' @param bar that
#' 
#' @examples
#' 1 + 1

a single query could match all #' <content> lines, and then each line of <content> could be collapsed together to form a single "document" to be reparsed by tree-sitter-roxygen2 that looks like this

@param foo this
@param bar that

@examples
1 + 1

At least, I think that is how it works.