fix shebang parsing

[m4rch3n1ng]

fixes #226 as outlined in #226 (comment)

rust currently only strips shebangs that are on the same line (i.e. not separated by a \n, see strip_shebang in the rustc_lexer), but the previous regex allowed newlines and didn't correctly parse empty shebangs or shebangs with only a / chracter.

now the regex explicitly disallows \n characters and parses until it finds a newline or fails to parse if it finds a [. this sadly also introduces a new small regression: tree-sitter can no longer parse rust files that are just a shebang and nothing else, no newline. i would like to fix it and the fix would be to replace the \n at the end of the regex with a (\n|$), but if i do that then tree-sitter will complain with Error processing rule shebang: Grammar error: Unexpected rule ExpandRegex(Assertion), so i just left it. i don't think that should be much of an issue probably.

technically speaking the rust lexer allows all of these characters as whitespace, but i thought the current selection of ascii whitespaces is probably enough.