Add option to compute FR filter bounds from m-mode spectra

[r-pascua]

With the push to a per-baseline style analysis, it would be helpful to integrate the code for determining fringe-rate filter parameters from $m$-mode power spectra into hera_cal. The basic idea is to allow the user to provide per-baseline $m$-mode power spectra, then compute main lobe filter parameters given a frequency taper, percentile cuts, and time information. The code should be implemented in a modular, generalized form.