Change identifier replacer to allow for all valid Python identifiers

[marvinvanaalst]

Overview

Change check for valid identifiers from

_IDENTIFIER_PATTERN: ClassVar[re.Pattern] = re.compile(r"^[A-Za-z_][A-Za-z0-9_]*$")

to using the stdlib Python functions str.isidentifier and keyword.iskeyword

Details

The previous identifier pattern has two problems.
a) It blocks a lot of names as a lot of UTF-8 characters, e.g. α , are valid identifiers
b) It doesn't block invalid keywords as Python names, e.g. def

The stdlib str.isidentifier and keyword.iskeyword functions allow to correctly check for both cases and as a bonus are maintained by the Python folks, so new keyword additions etc. aren't a problem

I replaced the check in the IdentifierReplacer class and added an additional check for the def keyword in the test cases.

As a side note: is it necessary to check for the key to be a valid Python identifier? Since the key comes from parsing an existing Python function it cannot be invalid in the first place

[marvinvanaalst]

Oh and optionally I'd change the value error message to something like f"'{v}' is not a valid Python identifier., because when I first read the error message I was a bit confused as to what exactly identifier name was referring to (e.g. LaTeX or Python or something internal to the package)

[odashi]

Great, thanks!