Fix _codecs.escape_decode return type

[ledvinap]

Summary

Fix the stdlib stub for _codecs.escape_decode so it returns tuple[bytes, int] instead of tuple[str, int].

Also add regression coverage in stdlib/@tests/test_cases/check_codecs.py to verify the inferred return type for codecs.escape_decode.

Details

Python's docs describe codecs.escape_decode as returning bytes, and CPython runtime behavior matches that. The old stub returned str, which then propagated
to codecs.escape_decode because codecs.pyi re-exports _codecs.

I kept the input type as str | ReadableBuffer rather than narrowing it to bytes-like only, because current CPython still accepts str input at runtime and
returns (bytes, int).

AI assistance

This PR was prepared by OpenAI Codex/GPT-5 and is human-validated.

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

mitmproxy (https://github.com/mitmproxy/mitmproxy)
+ mitmproxy/utils/strutils.py:125: error: Unused "type: ignore" comment  [unused-ignore]

[brianschubert]

LGTM, it seems that this was an edge case since the Python 2 days, where this function returned a str while the other *_decode functions returned a unicode object. See remark below about adding a test.

[brianschubert]

Test cases are only needed for weird and/or complicated cases that might regress in the future. These tests basically repeat the signature, so I don't think they're necessary

Suggested change

	assert_type(codecs.escape_decode("ab"), tuple[bytes, int])
	assert_type(codecs.escape_decode(b"ab"), tuple[bytes, int])
	decoded, consumed = codecs.escape_decode(b"ab")
	assert_type(decoded, bytes)
	assert_type(consumed, int)