HTTP redirects are ignored during git-fetch over HTTP

[rtyley]

Arnt Gulbrandsen reports that Agit fails to follow HTTP redirect codes when cloning - ie when a repo has moved. Points noted:

• doesn't clone or pull from the new location.
• doesn't ask whether the location might be wrong, or suggest pulling from the 301-specified location instead
• doesn't allow manually specifying a new pull location

Arnt supplied a sample repo which redirects: http://git.aox.org/aox.git → https://github.com/aox/aox.git

A quick test with CGit shows that it succeeds in cloning the example from the original git.aox.org source url - so it's probably hitting paths on the old url space and getting redirected for every hit that forms part of the 'fetch' operation. The git config file still shows a fetch url of http://git.aox.org/aox.git after the clone is complete.

Patching JGIt to get it's http client to follow redirects like CGit is an option... we should obviously allow the user to edit the urls for the remotes as well.