Speed up compressor tree generation

[hansemandse]

The current test flow takes beyond 10 minutes to complete a single full run mostly because of long execution times in the compressor tree generator. I believe the flow can be sped up, but it may require some timing analysis to better understand which part of the algorithm takes so long to execute.

[hansemandse]

After some initial experiments using System.nanoTime on parts of the generator's methods, it seems the largest part of the execution time stems from Chisel itself, not approx. My guess is that this is due to the resulting circuits making use only of pure logical building blocks rather than high-level arithmetic operators. Perhaps #5 will help to fix this?