[Yum] Faster downloads of headers

[seth vidal]

> 1. Use HTTP pipelining.  HTTP 1.1 allows sending multiple requests 
> before reading the responses.  I don't know if the Python httplib 
> supports pipelining.

keepalive is already implemented in urlgrabber.

> 2. Put the headers in an archive.  The initial fetch of a repository 
> would download the archive if present and unpack it.  Later updates 
> would fetch the individual headers.

the headers are going away in the next major version or so, so this is
sorta moot point.

-sv