Specify the bandpower definitions for discretization

[aewallwi]

We should discretize the bandpowers in a way consistent with assumptions in how the data is discretized. This may require adding a new field to uvpspec.

[aewallwi]

@acliu, you mentioned that our Q-matrices sampled at the centers of k-perp bins but integrated over k-parallel bins. It sounds like we might want to have different k-perp and k-parallel definitions.

We should write down explicitly how we would translate this to the spherical power spectrum.

Maybe we should consider addinge new fields added to uvpspec so that this information can be communicated to pspec_likelihood.