Improve Live FIFO-Sizing

[fpjentzsch]

Live FIFO-Sizing was first introduced here: #46

Some improvements remain:

• Implement a full binary search / bisection instead of the current two-phase algorithm
• Implement an optimized RTL implementation of the virtual FIFO. This could save LUTs (current HLS variant uses ~ 300 LUT per virtual FIFO) and reduce the FIFO depth offset and thus the minimal FIFO depth the algorithm is able to determine (currently 9, ideally 2 or even 0 (no FIFO needed at all))
• Switch to a better control mechanism, i.e., a ring/chain/stream based virtual FIFO configuration interface instead of AXI-Lite
• Support Alveo/Versal platforms in addition to Zynq