feat: add RFC 7616 support for non-Latin credentials in HTTPDigestAuth

[Jo2234]

Summary

Implement RFC 7616 extensions to fix HTTPDigestAuth failing with non-Latin-1 usernames (e.g., Cyrillic, Czech diacritics).

Problem

import requests
from requests.auth import HTTPDigestAuth

# This fails — 'ř' cannot be encoded as latin-1
auth = HTTPDigestAuth('Ondřej', 'heslíčko')
# The Digest header produces: username="Ondřej" which is invalid

# Workaround from issue produces garbled bytes:
auth = HTTPDigestAuth('Ondřej'.encode('utf-8'), 'heslíčko')
# Produces: username="b'Ond\\xc5\\x99ej'"  — wrong!

Solution

This PR implements three RFC 7616 features:

1. username* parameter (RFC 7616 §3.4 + RFC 5987)

When the username cannot be encoded as Latin-1, the username* parameter with RFC 5987 encoding is used:

# Before (broken):
Digest username="Ondřej", ...

# After (correct):
Digest username*=UTF-8''Ond%C5%99ej, ...

2. userhash support (RFC 7616 §3.4.4)

When the server sends userhash=true in the challenge, the username is hashed for privacy:

Digest username="<hash of username:realm>", ..., userhash=true

3. charset echo (RFC 7616 §3.3)

When the server advertises charset=UTF-8, the client echoes it back.

Changes

File	Change
src/requests/auth.py	Added _is_latin1_encodable(), _encode_rfc5987(), and updated build_digest_header() with RFC 7616 username handling
tests/test_digest_rfc7616.py	12 new tests covering: Latin-1 detection, RFC 5987 encoding, username* parameter, userhash, charset, and response digest validity

Backward Compatibility

• Latin-1 encodable usernames (ASCII, Western European): No change in behavior — the standard username="..." parameter is used as before
• Non-Latin-1 usernames (Cyrillic, CJK, etc.): Now work correctly instead of failing
• The response digest computation is unchanged — A1 = username:realm:password is still hashed with UTF-8

Fixes #6102