Adds default port handling to FormatUrl.

Adds default port handling to FormatUrl.

This is in preparation for making a "FormatSecurityOriginForTheUser" or
some such function; users don't care about the port if it is what they
expect such as 80/443 for http/https.

BUG=402698

[dewittj, 2014-08-22 01:19:38]

PTAL, this is my first time wading into net/ code so please let me know if I've
missed anything obvious.

Also the branching in there is horrible-looking, if you have a better suggestion
I'm all ears.

[dewittj, 2014-08-22 01:25:22]

probably need to add a test to FormatUrlParsed and FormatUrlWithOffsets too..

[dewittj, 2014-08-22 20:52:24]

This probably is defunct since GURL canonicalizes input urls, and part of
canonicalize removes the port if it's default:

https://code.google.com/p/chromium/codesearch#chromium/src/url/url_canon_etc.cc

line 200.

Let me know if you think there's still some utility for explicit removal of the
port in FormatUrl

[Ryan Sleevi, 2014-08-25 04:16:37]

So... did your unittests confirm it's not needed? :)

https://codereview.chromium.org/493253003/diff/1/net/base/net_util_icu.cc
File net/base/net_util_icu.cc (right):

https://codereview.chromium.org/493253003/diff/1/net/base/net_util_icu.cc#new...
net/base/net_util_icu.cc:755: if (!port_needs_reset && format_types &
kFormatUrlOmitDefaultPort) {
Put this bitmask in ()

(format_type & kFormatUrlOmitDefaultPort)

[dewittj, 2014-08-25 20:45:42]

Yes, unless there is any way to avoid the canonicalization step.  I think the
constructor
  GURL(std::string canonical_spec, const url::Parsed& parsed, bool is_valid);
could be used in this way but based on the name it would be improper usage.  I
think url_format doesn't need to worry about improperly constructed GURLs.

I'll reopen this if you disagree.