fixed DCM free vs idle discrepancy in multithread configuration

refactored association between processing threads and data source session to fix DCM free/idle discrepancy

For optimization reasons the number of event slots must be equal to the number of threads (one lock becomes unnecessary under this condition). It is still to be discussed if this is a acceptable constraint even in view of Phase 2. This is why this MR is marked as draft.

Validation was done in the testbed using pudummy as event processor: https://vm-atlas-tdaq-cc.cern.ch/tbed/pbeast/grafana/d/000000037/daq-backpressure?orgId=1&from=1689254695434&to=1689256208200 The run between 15:26 and 15:30 has the fix in; free and available cores from DCM are equal. The run between 15:36 and 15:43 is without the fix.

In pudummy the branch "multithread" now provides a real multithreaded implementation. pudummy and athena have very different multithreading strategy so testing with athena is necessary. I'm happy to provide support for this.

Relates to ADHI-4936.