Address ping back performance issues

CRM_Utils_VersionCheck is constructed I think for all requests. It uses cached data if available, and otherwise executes a request to latest.civicrm.org, with a 5 second timeout.

Just now latest.civicrm.org was down, and it's been down I think for most of the day. (I got hold of bgm in IRC, who restarted apache and brought it back up). The effect of this, since the cahce had expired was that every request was waiting for the 5 second timeout. This presumably affected every civicrm site.

I propose some combination of the mitigation approaches below:

• monitoring on latest.civicrm.org so it doesn't stay down long.
• move the pingBack into a cron job.
• Reduce the timeout to 1 second. I've done this locally as a quick and easy mitigation, but the cron approach would be a better alternative.
• if not using cron, then maybe delay the pingBack request till after the page content has been sent to the user?
• cache information about request failure to limit the frequency of re-tries.

I've put this at major priority, given significant impact on all requests for everyone, but given that the impact is slowing things down and everything still works eventually, that priority is at least arguable.