Fix #177643 by making pycurl handle url-embedded credentials again.
* bzrlib/tests/test_http.py:
Define proper tests for pycurl instead of replacing the urllib
ones (a correct refactoring is already under work elsewhere). Add
KnownFailures to document limitations in the pycurl
implementation (cant prompt for password, no support for
authentication.conf, cant handle a nonce change for digest auth).
(TestAuth.test_no_prompt_for_password_when_using_auth_config): Fix
typo.
* bzrlib/transport/http/__init__.py:
(HttpTransportBase._remote_path): Don't mention user and passwords
in urls since bothe implementations handle credentials separately.
(HttpTransportBase._create_auth): Share default credentials
creation between both implementation.
* bzrlib/transport/http/_pycurl.py:
(PyCurlTransport._get_curl): Save user provided credentials.
(PyCurlTransport._set_curl_options): Set USERPWD explicitely from
credentials if any have been provided in the url.
* bzrlib/transport/http/_urllib.py:
(HttpTransport_urllib._remote_path): Deleted, pushed to base class.
(HttpTransport_urllib._perform): Simplified.
* bzrlib/transport/http/_urllib2_wrappers.py:
(AbstractAuthHandler._build_password_prompt): Fix typo.