Better allocation of 4op channels

[jpcima]

If I believe the implementation of realTime_noteOn, the allocation of natural 4op is not really ideal.

What is happenning:

• select a first channel which is best scored for 2op
• select a second channel as first channel + 3

But really, there might exist a better choice of channel pairs if a different first channel was chosen. In this case, I would consider a following strategy for 4op: go over all candidate pairs, and consider for selection the pair which has the largest sum of scores of its 2 channels. EDIT rather, a computation of "secondary goodness" which takes at least in consideration the channel's off time

[Wohlstand]

The quote from killOrEvacuate:

 // FIXME: This does not care about four-op entanglements.

that was left by Joel in original ADLMIDI and was kept