To better understand how fast we are servicing Download page requests
we should consider logging the number of seconds from when a SavePageLater
request is made to when we first start processing it. Additionally we
might consider separate histograms for svelte and non-svelte devices
and the latter can immediately schedule processing. This would help
provide insight into the potential value or improvement of additional
scheduling approach (eg, ConnectivityListener-based) while Chrome is
resident on the device.

[Note: We do already have UMA for the time to when the Save completes
so this is more specifically about the initial scheduling.]