fix: decode percent-encoded credentials from URL

[NotFoundOvO]

Summary

When credentials are percent-encoded in the URL (e.g. http://u%40d:***@host/), Python's urlsplit() preserves the percent-encoding in .username and .password attributes. This causes HTTPie to send the raw encoded string (u%40d) as the username instead of the decoded value (u@d).

This PR applies urllib.parse.unquote() to decode URL credentials before passing them to the auth handler, matching curl's behavior.

Changes

• Import unquote from urllib.parse in argparser.py
• Apply unquote() to url.username and url.password in _process_auth()

Reproduction

# Before fix: returns 401 (wrong credentials)
http https://u%40d:***@httpbin.org/basic-auth/u%40d/1%3d2%3f

# After fix: returns 200 (credentials correctly decoded)
http https://u%40d:***@httpbin.org/basic-auth/u%40d/1%3d2%3f

Closes #1623

[NotFoundOvO]

Hi, I noticed the CI test failures across all platforms. After checking, the same failures (Tests + Coverage) are also present on the branch — this appears to be a pre-existing CI infrastructure issue, not related to this change.

The code change itself (using to decode percent-encoded URL credentials) is straightforward and correct. does not auto-decode percent-encoded username/password, so the explicit is needed.

Would appreciate it if someone could take a look when the CI is fixed. Thanks!