Return None instead of crashing on invalid An+B input

[gaoflow]

Steps

tinycss2.nth.parse_nth is documented to return None for invalid input, but several short, plausible truncated An+B fragments crash instead:

from tinycss2.nth import parse_nth
parse_nth("+")     # StopIteration
parse_nth("+/**/") # StopIteration
parse_nth("n+")    # AttributeError: 'NoneType' object has no attribute 'type'
parse_nth("2n +")  # AttributeError

Two unguarded token reads:

• parse_nth (after a leading +): next(tokens) reads the next token without guarding against an exhausted iterator. The raw next is intentional — whitespace after + must stay invalid, so it can't be skipped — but exhaustion wasn't handled.
• parse_signless_b: _next_significant(tokens) can return None, then token.type dereferences it.

Fix

Use next(tokens, None) and add token is not None guards so both paths fall through to the function's implicit return None, matching the documented contract.

Whitespace-sensitivity is preserved: +n → (1, 0) (valid), + n → None (invalid), confirmed against the official An+B.json test data.

Added regression tests in tests/test_tinycss2.py; the full suite passes (16975).

Notes

Found by fuzzing the public API. This pull request was prepared with the assistance of AI, under my direction and review.

[liZe]

Thank you!