native parser: tokenizer error: <U+2EBF0> is not a valid identifier

[jakkdl]

Hypothesis found a failing test case in the flake8-async CI which boiled down to libcst's native parser incorrectly marking 𮯰 as an invalid identifier. It is a valid identifier, but only on 3.13+. The pure parser works as expected.

Repro:

import libcst
libcst.parse_module('class 𮯰: pass\n')

$ export LIBCST_PARSER_TYPE=native
$ python repro.py
Traceback (most recent call last):
  File "/tmp/f8as_crash/foo.py", line 2, in <module>
    libcst.parse_module('class 𮯰: pass\n')
    ~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^
  File "/tmp/f8as_crash/.venv/lib/python3.13/site-packages/libcst/_parser/entrypoints.py", line 109, in parse_module
    result = _parse(
        "file_input",
    ...<3 lines>...
        detect_default_newline=True,
    )
  File "/tmp/f8as_crash/.venv/lib/python3.13/site-packages/libcst/_parser/entrypoints.py", line 55, in _parse
    return parse(source_str)
libcst._exceptions.ParserSyntaxError: Syntax Error @ 1:1.
tokenizer error: "𮯰" is not a valid identifier

class 𮯰: pass
^

$ python --version
Python 3.13.1
$ pip list | grep libcst
libcst                1.6.0