refactor(ephemeris): Use ch_ephem

[ketiltrout]

This re-imports stuff moved to ch_ephem. It both re-imports everything that was moved back into ch_util.ephemeris for backwards compatibility, but also removes references to ch_util.epheremis from the code, using the newly-exported versions of things directly.

This PR requires both:

• https://github.com/radiocosmology/caput/pull/271
• https://github.com/chime-experiment/ch_ephem/pull/1

Catalogs:

All the catalogs have been moved to ch_ephem, which provides a standard interface to them via ch_ephem.catalogs.load. To make this change, fluxcat.FluxCatalog.load has been broken into two pieces. The new fluxcat.FluxCatalog.load_dict imports a catalog from an already-parsed JSON dict (which is what ch_ephem provides). The original fluxcat.FluxCatalog.load works as it used to, but tail-calls load_dict to do most of the importing.

Position data:

Position data (lat, lon, alt) is now obtained by fetching the parameter out of the caput.time.Observer objects provided by ch_ephem. The constants in ch_util.ephemeris are populated by doing this.

Instrument rotation:

ch_ephem has also taken on the rotation data that was in tools (where it's been re-imported from ch_ephem). Not sure this is ideal. Please comment on the ch_ephem PR if you can provide an opinion on this.

Private objects deleted:

I've also deleted a number of private objects which are no longer being used (i.e. objects whose names start with a leading underscore).

Notable among these is the unused ch_util.tools._LAT_LON which had position data for various instruments which differed from the position data provided by ch_util.ephemeris.

Also removed are the long-deprectated ch_util.ephemeris._get_chime and ch_util.ephemeris.transit_RA.

[ketiltrout]

The checks won't work until the other PRs are merged.

[rikvl]

I had a look at moving the HFB catalog. That looks fine to me. Apologies for the delay!