[Test] Add dataflow partitioning, folding, ip gen, ip stitch and first draft of verification of data flow part

this is to provide the equivalent "programmable full" behavior
with cap 32 threshold 16, as per recommendation by Lucian P

this is essential for the streaming weights to work correctly
due to the depth of the mem addr gen pipeline

turns out HLS does not let us specify stream depths on top level
AXI streams, need another solution