Lingering wish list:

- parameterized spectral behavior (1/nu, tabular?)

- build array for a specific WCS (deal with rotated/weird images)

- instantiate from CASA header

- compare two beams (or an array of beams) and recommend a "common
  beam" to target for a matched resolution image. Analytic solution
  and then tolerance ~ Nyquist on top of the analytic solution.

- expose the factor (area ratio) to normalize by in convolution in
  order to keep Jy/beam (i.e., to use the new beam as a unit)
