Adapt functionality of Classad-Scheduler

[eileen-kuehn]

The workflow should be as follows:

• Perform autoclustering of jobs,
• go through job list independent of auto cluster,
• assign job to one of the machines by going through machine cluster,
• remember assigned machine cluster for the job cluster the job belongs to,
• when looking for another job, check if it is in an auto cluster where we already know which machine cluster is good; if existing, directly use it and don't search for another matching machine cluster.