When a target's `last_perf_capture_on` is far into the past, the next
attempt to capture metrics for that target will result in an attempt to
gather a potentially huge amount of metrics at one time.

This change will break down the intervals between `last_perf_capture_on`
and `today` into individual days.  This is the same logic that's used
when collecting gap metrics.

https://bugzilla.redhat.com/show_bug.cgi?id=1429593

Remove a redundant call to `map`, and add a test to make sure that the
dates in metrics collection are in a properly reversed order.

https://bugzilla.redhat.com/show_bug.cgi?id=1429593

Ensure that perf_capture_queue works on long intervals and single days.