fix: use urlsplit instead of urlparse for digest auth URI

[veeceey]

Summary

• Fixes the uri field in HTTP Digest Authentication being incorrectly truncated when the URL path contains semicolons
• Replaces urlparse with urlsplit in build_digest_header() since urlparse treats semicolons as parameter separators (stripping everything after the first ; from the path), while urlsplit correctly preserves them

Fixes #6990

Explanation

Python's urlparse splits the path on ; and moves the content after it to a params attribute:

>>> from urllib.parse import urlparse
>>> urlparse("https://example.com/a;b?q=1")
ParseResult(scheme='https', netloc='example.com', path='/a', params='b', query='q=1', fragment='')

urlsplit does not have this behavior and keeps the full path intact:

>>> from urllib.parse import urlsplit
>>> urlsplit("https://example.com/a;b?q=1")
SplitResult(scheme='https', netloc='example.com', path='/a;b', query='q=1', fragment='')

Since only path and query are used from the parsed result (not params), urlsplit is a safe drop-in replacement that correctly preserves the full request URI as required by RFC 7616.

Test plan

• Verified the fix with the exact reproduction case from the issue (MusicBrainz URL with semicolons as separators)
• All existing non-httpbin tests pass

[veeceey]

Heads up: this PR currently shows as behind the base branch (merge conflict likely). I'll rebase against main to resolve this and keep it mergeable.