Only calculate features for instances before cutoff

[CJStadler]

When a single cutoff_time is provided.

Resolves #437

[codecov]

Codecov Report

Merging #523 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master    #523      +/-   ##
=========================================
+ Coverage    96.1%   96.1%   +<.01%     
=========================================
  Files         108     108              
  Lines        8898    8913      +15     
=========================================
+ Hits         8551    8566      +15     
  Misses        347     347

Impacted Files	Coverage Δ
...imitive_tests/test_groupby_transform_primitives.py	100% <100%> (ø)	⬆️
...computational_backends/calculate_feature_matrix.py	97.09% <100%> (+0.01%)	⬆️
...utational_backend/test_calculate_feature_matrix.py	99.33% <100%> (+0.01%)	⬆️
featuretools/tests/dfs_tests/test_dfs_method.py	98.41% <100%> (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 4cbca08...fd9c909. Read the comment docs.

[rwedge]

From the coverage report it looks like some code related to missing data relied on the old way of using all the instances in order to be tested.

https://github.com/Featuretools/featuretools/blob/269307e8b1e2214e8ad422d524e412ce0d47f4b3/featuretools/computational_backends/pandas_backend.py#L252-L255

https://github.com/Featuretools/featuretools/blob/269307e8b1e2214e8ad422d524e412ce0d47f4b3/featuretools/computational_backends/pandas_backend.py#L277-L280

We might be able to test this code with a test using a single cutoff time and a list of instance ids that includes instances whose time index is after the cutoff time.

[CJStadler]

@rwedge I added a test and I think the coverage is good now. Back to you.

[rwedge]

Looks good