Frequency Bands Class to support non-contiguous intervals

[kkappler]

Not high priority
FrequencyBands are currently specified by a lower_bound and an upper_bound. This handles most cases but it is not unrealistic that we may encounter bands which have a few contaminated harmonics and we would like to suppress these. There is currently no method for handling this.

Of course we could split a given band into a sequence of contiguous sub-bands, but in that case we have fewer observations to input to robust regression. This could compromise the quality of some estimates.

It would be nice if a FrequencyBand object could specify some harmonics to ignore.

This of course breaks compatibility with EMTF z-files but those are slated for deprecation in the long term anyhow

[kkappler]

This maybe more more trouble to implement than it's worth. Needs review and prioritization, and if we address it, it needs definition of use cases.

Also, this actually depends on issue #150 being fixed first, since currently, EMTF band setup is still baked into aurora.

[kkappler]

When reviewing this, also consider:

• Add band_edge_definition = ["exact", "canonical"] to config
• Definition of some bands templates that people can use
• logarithmic spaced band templates

[kkappler]

This would be better handled by using masks in weighting functions (#61). Marking this as closed