Vectorize mpc.mpc_orbit, mpc.comet_orbit

[JoshPaterson]

This commit is only changes that are necessary to vectorize the creation of _KelperOrbit objects. This allows the at() method to work but not the .observe method. This PR needs to be merged before #526.

In order for the tests to pass I need to change them so that they don't use VectorSum objects or the observe method.

[xmichaelx]

I see that this and #526 got stuck - is there a way I could help them go through?

[xmichaelx]

@brandon-rhodes I've created PR to @JoshPaterson fork with changes allowing for observe.at to work. One weird issue is that it seems like rows with "E" in uncertainty column screw up propagation for all orbits when passed to mpc.mpcorb_orbit, because when I filtered them out everything worked perfectly ok.

Filtering out done using:
mpcorb_data = mpcorb_data[mpcorb_data.uncertainty != "E"]

Other than changes in PR not looking pretty it seems that it should work. Although some testing and refactoring would definetely not hurt.

[brandon-rhodes]

Thanks for the update! I have some time this week for open source work, so I'll expect to prioritize this parallel-processing Skkyfield issue. Thanks for reviving it.

[xmichaelx]

Awesome, there's somewhat related issue for using data form mpcorb.dat files #650 - found out about while trying get precise correlation with telescope observations and comparing differences with orbits pulled from JPL Horizons. That one fortunately would only require adding a paragraph or two to chapter on Kepler orbits as a warning.

[JoshPaterson]

Now that it's getting cold outside I'll have time to devote to this as well!

@xmichaelx It'll probably take me a little bit to get back up to speed, then I'll get your PR merged, thanks for making that!

[jurezakrajsek]

Hi @brandon-rhodes, @JoshPaterson

Is there any progress on this topic of vectorizing the comet objects to enable vectorized calculations on multiple comet objects simultaneously?
I would like to switch from the pyephem library to SkyField in my COBS project, where I currently use pyephem to calculate the visibility of comets for a given time frame and user location.
I have approximately 3000 comets in the database, and using a for loop to calculate them one by one with SkyField is too slow for this to be usable as a real-time function on the website.

Best regards
Jure