In a combination of circumstances, n_coarse_chan is incorrect

[texadactyl]

Context: data_handler.py, function __split_h5().
Eyeballed. I am not sure if this would come up naturally.
Also, adding data_handler logging to let the operator know if the n_coarse_chan field
is supplied from the operator, computed by blimpy, or or computed based on the header nfpc field.

nfpc :: Number of fine channels per coarse channel.
n_coarse_chan = Number of fine channels / nfpc.

Fix and retest with a new .h5 file from rawspec.