Improve memory usage through support for pandas categorical types

[kmax12]

Currently, we convert columns that are pandas categorical types to an object type. This is bad for memory usage.

This pull request changes removes that conversion so that the categorical dtype is preserved while running featuretools.

Test cases pass, but more integration and performance testing is needed.

[codecov-io]

Codecov Report

Merging #196 into master will increase coverage by 0.05%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #196      +/-   ##
==========================================
+ Coverage   93.21%   93.26%   +0.05%     
==========================================
  Files          70       70              
  Lines        7727     7726       -1     
==========================================
+ Hits         7203     7206       +3     
+ Misses        524      520       -4

Impacted Files	Coverage Δ
featuretools/primitives/cum_transform_feature.py	97.63% <100%> (ø)	⬆️
...utational_backend/test_calculate_feature_matrix.py	99.25% <100%> (ø)	⬆️
...computational_backends/calculate_feature_matrix.py	97.02% <100%> (+0.03%)	⬆️
featuretools/tests/testing_utils/mock_ds.py	87.4% <100%> (+0.18%)	⬆️
...turetools/computational_backends/pandas_backend.py	91.29% <100%> (ø)	⬆️
featuretools/entityset/entityset.py	94.59% <100%> (+0.77%)	⬆️
featuretools/utils/gen_utils.py	64.44% <0%> (-2.23%)	⬇️

---

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 ee25b59...4c0513a. Read the comment docs.