fastesturl doesn't reflect real bandwidth

[Jerry Zhao]

Originally, fastesturl function use curl to download the root path of http server, but the response time doesn't reflect the real download speed of the file.

Propose using the download time for the first chunk of 20000 bytes of the file to determine the fastest server.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: dev/experimental
Review: https://review.openstack.org/134611

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/134611
Committed: https://git.openstack.org/cgit/stackforge/compass-core/commit/?id=74b5b57e03d66dade5034c8d0082c2ab9bb50c32
Submitter: Jenkins
Branch: dev/experimental

commit 74b5b57e03d66dade5034c8d0082c2ab9bb50c32
Author: zhaoxinyu <email address hidden>
Date: Fri Nov 14 09:53:29 2014 -0800

    Use download speed to decide which url is faster

    just getting the root path of a http server to measure the response
    time does not necessarily reflect the real download speed of a source.
    reponse can be 302 http redirect which may be really fast but download
    bandwidth is determined by the route to the real source http server.
    Therefore, use download time for the first chunk of 20000 bytes to
    determine which is faster.

    Closes-bug: #1392816
    Change-Id: I3019dce3cf92dd75c58a21987c653cbcb502264b