URL constructor drops final-segment semicolon parameters

[AryanRG]

Constructing a URL from a complete string drops valid final-segment path parameters before a request is sent:

>>> httpcore.URL("http://example.com/items;version=2?x=1").target
b'/items?x=1'

The expected target is b'/items;version=2?x=1'. The URL constructor uses urllib.parse.urlparse() but assembles the target without parsed.params; urlsplit() retains the complete path. This also affects an empty trailing semicolon (/items;). Explicitly constructed URL(..., target=...) values already preserve it.

Local reproduction on the current master head (10a658221deb38a4c5b16db55ab554b0bf731707) confirmed the altered request target with sync and async HTTP/1.1 requests, direct and through a forward proxy. A focused one-line parser change and regression tests pass the local suite and scripts/check (230 tests passed, with six pre-existing XPASS results). No PR has been opened.

Is retaining those semicolon path parameters the intended behavior, and would a focused patch be welcome?