Raise a clear TypeError for non-str/non-bytes input to slugify()

[rantkiglorelgmx-lab]

What

slugify() currently fails with an opaque error for input that is neither str nor bytes (e.g. None, int) — the decode path raises TypeError: decoding to str: need a bytes-like object, NoneType found, which doesn't clearly say what's wrong. This adds an explicit, clear guard at the top of the function:

if not isinstance(text, (str, bytes)):
    raise TypeError(f'text must be str or bytes, not {type(text).__name__}')

Backward compatibility

Add-only — every input that previously worked is unaffected. In particular, bytes input is still decoded exactly as before (slugify(b'Hello World') == 'hello-world', confirmed unchanged). Only the message for already-erroring input (non-str/non-bytes) is clearer.

Verification

• New test asserts the clear message via assertRaisesRegex, and separately asserts bytes input still slugifies correctly (regression lock).
• Non-vacuity: reverted the source fix only — the new test fails on the old code (message mismatch), confirming it isn't a vacuous test.
• Full suite green: 83 passed.

Co-authored by Claude (Anthropic) as part of an automated hardening pass; independently re-verified (non-vacuity, backward-compatibility, coverage) before submission.

[jquast]

But the function is typed already and would be caught by type checkers.

If a developer (or AI bot) is having trouble with "opaque errors" because of using the wrong type, then they could use a type checker themselves, instead of depending on all upstream dependencies to add if isinstance() checks to every function, this could go on forever, billions of lines of code to add to all public api functions of all popular python projects that negatively impact performance of all users for the very small benefit of a few confused developers.