XrdCl::URL accepts invalid URL and rejects valid ones as invalid

[amadio]

While converting the XrdCl::URL tests to GoogleTest, I noticed that it accepts zero, negative, and big (>65536) port numbers as valid. For example, "root://eospilot.cern.ch:-1", and "root://eospilot.cern.ch:281474976710656" are both considered valid. I also noticed that a URL with parameters, but without a path will be considered invalid, i.e. something like http://localhost:8080?param=value. However, that should likely be considered valid, with path normalized to "/", as that's what the browser does. URLs with empty username and/or password are also rejected, i.e., something like http://:@localhost and http://@localhost. Should these be considered valid? The browser seems to accept them. The final thing is that XrdCl::URL cannot accept passwords that contain the character @ or / in them, but I guess that's not really a bug, even if those are supposedly valid characters for passwords.

[bbockelm]

Per https://www.rfc-editor.org/rfc/rfc1738, it appears that the password should be escaped in order to allow reserved symbols like @.

Note the interesting semantics to differentiate between “no user or password is set” and “user and password are set to an empty string”.

[amadio]

Thanks for the link to the RFC. I think we should try to follow it as closely as possible.