Fix unicode-named data loading

[mraspaud]

Dataset loading was failing in python 2 if the query was done by name with a unicode name.

• Tests added
• Tests passed
• Passes git diff origin/develop **/*py | flake8 --diff

[djhoese]

In my opinion this is not the right fix for this. This bug that is being encountered is an issue in the readers trying to use unicode objects (in python 2) as normal string values. Your change in this PR is making satpy support unicode objects where it should only support str objects.

I think an "easy" fix would be to modify the yaml reader to convert ancillary metadata to strings before trying to load it. The "best" solution is really to modify the ancillary data stored in the .attrs that the reader is returning to make them not unicode objects. This may be difficult for generic readers that just copy the .attrs from the NetCDF file to the produced dataset.

I am ok with this change if time is an issue, but we could also just suggest moving to python 3 😉.

[djhoese]

My argument is shutdown by the fact that my_dict[u'my_key'] works for str keys.

[coveralls]

Coverage remained the same at 65.52% when pulling 1b8b2ce on fix-ancillary-data-loading into 43475c7 on develop.

[codecov]

Codecov Report

Merging #290 into develop will not change coverage.
The diff coverage is 66.66%.

@@           Coverage Diff            @@
##           develop     #290   +/-   ##
========================================
  Coverage    65.52%   65.52%           
========================================
  Files          117      117           
  Lines        14275    14275           
========================================
  Hits          9353     9353           
  Misses        4922     4922

Impacted Files	Coverage Δ
satpy/readers/__init__.py	94.75% <66.66%> (ø)	⬆️

---

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 43475c7...1b8b2ce. Read the comment docs.