Add per-observation MJDs [ADAM-70]

[moeyensj]

To address the LSST use case, this PR adds per-observation MJDs to the binary files. In instances where the observation is not measured at the midpoint of the exposure then a 2-body propagation is carried out to update the predicted location of the orbit.

Note:

• This PR removes the approximately_propagate and the linear propagation functions in favor of making 2-body the standard for approximate propagations
• The tolerance for the synthetic indexing test has been increased to account for errors introduced by approximate propagation (tests are now within a milliarcsecond: this is still 10-100x larger than the errors we expect in astrometry from LSST)
• "mjd_start", "mjd_mid" are renamed to "exposure_mjd_start", "exposure_mjd_mid" internally
• For inputs to the indexing code and for the results returned by precovery "_utc" is appended the above to explicitly state the time scale (so "exposure_mjd_start_utc", "exposure_mjd_mid_utc")

[moeyensj]

@spenczar If I did the rebase correctly, I think I addressed all your comments. Ready for another pass!