[FanOutOptimization] Improvements to support LUT replication and more

[clavin-xlnx]

This improves FanOutOptimization in the following ways:

1. Instead of only supporting nets that are driven by flops, this also supports nets driven by LUTs and will replicate them accordingly.
2. Adds a flag that enables replicated cells (and the original cell) to be placed in partially used sites. Previously, this was problematic, but the bugs should be fixed now. However, when running with this option, users will probably need to run route_design instead of route_design -preserve.
3. The clustering is now done before cell replication and cells are placed based on their cluster sinks. If a LUT is being duplicated, the centroid for placement also includes the source drivers of the LUT inputs as they are part of the timing path.