Add ability to use dask distributed when generating animation videos

[djhoese]

This PR is a rewrite of the animation saving logic in the MultiScene. It removes the "cascaded" dask graph that I was using before because I discovered this was not multiprocess safe (opening a new ffmpeg connection on every process/node). I now give the option to use distributed to upload frames to be computed and write them as they complete. My hope is that some users with higher performance systems (or clusters) can take advantage of their systems specs by specifying batch_size and computing multiple frames at a time.

If a user does not have distributed installed then a basic one-frame-at-a-time computation will be done using normal dask computations.

• Tests added
• Tests passed
• Passes git diff origin/master -- "*py" | flake8 --diff
• Fully documented

I should note that this is documented in the function but not emphasized in the sphinx docs. I'll consider adding that soon.

[codecov]

Codecov Report

Merging #609 into master will decrease coverage by <.01%.
The diff coverage is 82.22%.

@@            Coverage Diff             @@
##           master     #609      +/-   ##
==========================================
- Coverage   78.14%   78.13%   -0.01%     
==========================================
  Files         136      136              
  Lines       19555    19615      +60     
==========================================
+ Hits        15281    15327      +46     
- Misses       4274     4288      +14

Impacted Files	Coverage Δ
satpy/tests/test_multiscene.py	96.49% <100%> (+0.62%)	⬆️
satpy/multiscene.py	80.08% <74.19%> (-3.41%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7c264ee...24205cf. Read the comment docs.

[coveralls]

Coverage decreased (-0.005%) to 78.139% when pulling 24205cf on djhoese:feature-distributed-animations into 7c264ee on pytroll:master.