[WIP] merge yt-4 with master

[ngoldbaum]

This tracks merging with yt-4 branch with master. This is not going to be useful for review but I'm opening now so I can work on getting the tests passing.

We will have a discussion on the mailing list about merging this.

[ngoldbaum]

The py2 build failures can be ignored because master won't support python2 anymore once this PR is merged.

[matthewturk]

I've updated the merge branch with most recent changes on both master and yt-4.0. The vector operations were a bit tricky to merge, but I think I got it.

@ngoldbaum I'm not sure if we can edit the pull request, but if the messages get too annoying we can switch it to being from my fork/branch.

[ngoldbaum]

No worries, let me know if you need a hand with anything.

[matthewturk]

Sneaky proj and quad_proj change now fixed for the ytdata tests!

[matthewturk]

@qobilidop Not sure if you're the right person to ask, but you might have an idea. The answer tests for tipsy are failing, and I'm not sure why they were ever passing before, but you might know why!

Right now the Tipsy tests run:

for test in sph_answer(ds, 'galaxy.00300', 315372, tg_fields):
    ...
    yield test

We're testing these fields:

tg_fields = OrderedDict(                                                                                                                      
    [                                                                                                                                         
        (('gas', 'density'), None),                                                                                                           
        (('gas', 'temperature'), None),                                                                                                       
        (('gas', 'temperature'), ('gas', 'density')),                                                                                         
        (('gas', 'velocity_magnitude'), None),                                                                                                
        (('gas', 'Fe_fraction'), None),                                                                                                       
        (('Stars', 'Metals'), None),                                                                                                          
    ]                                                                                                                                         
)  

The thing is, it seems to me that this should not actually ever pass -- the only fields that are of that particular size are the ones that are the all particle union.

So I guess I'm not sure why this ever worked? But more to the point, I wonder if you have any insight into why the unions and whatnot might give different results now.

[matthewturk]

Never mind, I figured it out! We now have an "nbody" type. I'll address this.

[jzuhone]

@matthewturk keep in mind that PR #2208 needs to be merged before a lot of these tests even work?

[matthewturk]

@jzuhone well crumbs, now I feel extra silly.

[matthewturk]

So #2208 has been merged, and I am going to poke more at this.

[matthewturk]

I'm closing and re-issuing from my repository!