Using libcurl to parse url

[fabricereix]

Currently the url crate is used to parse url in a hurl file, for example, to extract the host part. Libcurl could be used for that parsing, that would reduce the number of dependencies.

Besides, it seems that the url crate tries to resolve the host at parsing. Url::parse("toto") only succeed when toto is resolved (for example defined in /etc/hosts). The parsing should succeed without resolving the host.