Refactor delay-space measurement code and add features

[sjforeman]

This is meant to supersede #176 and #216:

• Refactors DelaySpectrumEstimator, DelaySpectrumEstimatorBase, and DelaySpectrumWienerBase to eliminate repeated code blocks and enable future extensions
• Implements the fixes and features from Misc. improvements to delay power spectrum estimator #216

Notably, I've kept the scheme of compressing many axes into a baseline axis. I don't think this is ideal in the long run, but in the short term, I think it's higher-priority to get the Wiener filter and #216 features all merged coherently

[ljgray]

This looks good to me. I have a couple of nitpicky stylistic comments but that's about it

[ljgray]

These changes look good to me overall. I spoke to @jrs65 on Friday and he mentioned having some thoughts about modifying the delay spectrum implementation to be split into multiple pipeline tasks. So I don't have anything else to change here, but we should let him weight in as well.

[Arnab-half-blood-prince]

Hi, I am also checking this and it looks good to me. However, I have one suggestion to add the freq axis in the index_map of the DelaySpectrum and DelayTransform container. This is needed to estimate various cosmological quantities downstream. For example, we need redshift to estimate comoving distances to get the kperp and kpar axes.

Probably something like this within the task -

delay_spec.create_index_map("freq",ss.freq)

And add this in the container -

@property
def freq(self):
    return self.index_map["freq"]