feat: strict encode, lenient decode for JWT subject claim

[gshmu]

Apply "strict with self, lenient with others" philosophy to JWT subject validation:

• Encode (strict): Validate sub must be string, raise TypeError if not
• Decode (lenient): Remove string type validation for sub claim in _validate_sub
• Tests: Update test to expect TypeError during encoding instead of decode error

Philosophy

This change embodies the principle of "严于律己，宽以待人" (be strict with yourself, lenient with others):

• Strict encoding: Ensures PyJWT generates standards-compliant tokens
• Lenient decoding: Allows PyJWT to consume tokens from other systems that may use non-string subject identifiers

[Copilot]

Pull Request Overview

This PR implements a "strict encode, lenient decode" approach for JWT subject claim validation. When encoding JWTs, PyJWT will strictly validate that the sub claim is a string and raise a TypeError if not. When decoding JWTs, PyJWT will be more lenient and accept non-string subject identifiers from other systems.

• Adds string validation for sub claim during JWT encoding with TypeError for non-strings
• Removes string type validation for sub claim during JWT decoding
• Updates test to verify encoding validation instead of decode validation

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
jwt/api_jwt.py	Adds sub string validation in encode method and removes string validation from decode's _validate_sub
tests/test_api_jwt.py	Updates test to verify TypeError during encoding instead of InvalidSubjectError during decoding