Enhancement of scheduler

[eileen-kuehn]

The scheduler currently assigns jobs to drones. If the starting of a job takes too long on a given drone it can happen, that another job is assigned to the very same drone. Therefore, the resources of a drone can exceed although the scheduler evaluated the job as fitting.

There seem to be two strategies on how to deal with this:

• the scheduler can assume that the resources provided by the drone are immediately reserved by the job so that only potentially matching jobs are send there (see PR Temporary update of available resources for drones #46), or
• the scheduler can push back the drone until the next scheduling interval.

Both modes should be supported.
I still need to check the HTCondor documentation to determine the default mode the scheduler should be running in.

This issue can help fixing the issue described in #39.