…t ends in the past

…#16)

These suggestions for [650 on the interuss
repo](interuss#650) primarily:

* Adjust the datetime comparison approach -- we don't want to mess with
time zones, or generally Python's suboptimal time treatment; we can
leave that to the `arrow` module we already import and use for many
other similar things
* Expand and make public the function to compute an end time of volume
or volumes
* Change StartOfTestRun to StartOfScenario in most test data -- I think
this is the most likely thing to fix most of the problems as it
decouples timing for a scenario from the rest of the scenarios in a test
suite/run