This commit adds in a number of planned features before the first big
`brutus` production run in the upcoming week over ~80M objects (!!!).
A summary of changes is below. Note that this commit also includes a number
of updates to the datafiles needed to run the code in the Dropbox folder
(linked through the README), which should hopefully finish uploading
sometime during the week.

Clusters
--------
- The baseline isochrones used to generate these had an error that caused
weird gaps in coverage. I think I've resolved this, and the isochrone
file has been updated accordingly.
- The `clusters` module now contains a pretty hefty log-likelihood function
that can be used to fit isochrones to co-spatial, co-eval populations of
stars. I don't have any documentation of this quite yet,
but I've confirmed in testing that this does actually work! I want to
build this out in the future, but for now I just wanted to push this
since it resolves #26.

MIST Models
-----------
- Using a series of ~solar and sub-solar benchmark clusters, I have derived
  improved corrections to the underlying SEDs on the lower-MS and associated
  zero-points. These are now implemented natively in the various SED generation
  tools by default, and so do not necessitate any more additional correction
  files. This resolves #27.

Binaries
--------
- I have improved flexibility for modeling binaries when generating individual
  SEDs and isochrones. These now allow users to specify where binaries are
  allowed and expand the default range.

Priors
------
- Added in the new 3-D dust map from Bayestar19, which is substantially higher
  resolution than Bayestar17 and looks very nice.
- Modified the default behavior of the dust prior based on the results from
  Green et al. (2019) so that we are now by default only adding scatter, not
  multiplying the existing scatter.
- Added support for an age prior that whose behavior tracks our
  metallicity priors in the disk and the halo.

Per-Object Fitting
------------------
- Cleaned up internals so things are a bit better organized (at least for me).
- Modified defaults to be more aggressive on adaptive thresholding
  and save fewer samples to improve runtime.
- Changed how parallaxes are used so they behave more like an additional
  "band" wherever possible, rather than being seen as a "special" additional
  piece of information. This is propagated through to various parts of the
  code and to the saved quantities (including the effective `Nbands`).
- Modified outputs so that names now are easier to understand.
- Added log-posteriors to outputs (both per-object and for individual draws).
  **THIS TOTALLY BREAKS BACKWARDS COMPATIBILITY.** I don't expect to have to
  substantially change the output format again, but I guess it's stil below v1
  for a reason!

Finally, here's a list of upcoming updates in the following week:
- A new model grid and zero-point file will be uploaded
  after updated Gaia passbands are added in and I can run additional
  systematics tests on high-latitude, low-reddening fields.
- Hopefully I can start trying to put together documentation (see #29) since
  there is now enough features in the code that people who aren't me will
  probably struggle even with all the excessive docstrings I add in.