Rewrite model_cmd to take a single method for errors and completeness

[cgarling]

A more complete description of observational errors and completeness requires all magnitudes at once, rather than the per-magnitude implementation we have currently. It is meant to be a simple implementation, but this generalization could make it much more useful. Then the case where the provided errfuncs and completefuncs are vectors could be handled by a special case method using an alternate dispatch signature.

[cgarling]

More useful would probably be a second call signature with the above implementation so that we could support both.

[cgarling]

Might consider adding photometric bias as well.