The rtbf is generally flexible to having the pfb integrated but it
undermines much of the way it was implemented (e.g., the way the post
beamform kernel was setup and switching `d_output` between
`d_beamformed` and `d_post_beamform`). However, this should still work.

The thinking is that for right now it is more important to show that the
PFB can work with the beamformer and to have a base benchmark on
profiling. I do not exactly need it in hashpipe yet. In the current
form this would work except the way the rtbf builds the pfb uses a
subproject and it is not allowed to have subprojects that have
subprojects and so for this to work with the xb-engine/pipeline the use
of the pfb in the rtbf this way would need to be able to access the pfb
as a subproject of xb-engine that lives next to rtbf (the other
subproject of the xb-engine)

But getting back to the idea; implement now beamformer+pfb to benchmark,
go work on other things outside the context of...