Optimize metadata calculation

[kmax12]

Currently, we recalculate the metadata for an entityset every time the property is accessed. This slows done many operations in Featuretools.

This pull request refactors things so we only calculate the metadata if it hasn't been created yet or if an operation invalidated it.

Locally this change makes the tests run in 134 seconds vs 513 without this change.

[codecov-io]

Codecov Report

Merging #229 into master will increase coverage by 0.03%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #229      +/-   ##
==========================================
+ Coverage   93.58%   93.61%   +0.03%     
==========================================
  Files          71       71              
  Lines        7727     7766      +39     
==========================================
+ Hits         7231     7270      +39     
  Misses        496      496

Impacted Files	Coverage Δ
featuretools/entityset/entityset.py	94.35% <100%> (+0.05%)	⬆️
featuretools/entityset/entity.py	88.86% <100%> (+0.05%)	⬆️
featuretools/tests/entityset_tests/test_es.py	99.35% <100%> (+0.04%)	⬆️

---

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 4970660...00518fa. Read the comment docs.

[rwedge]

Wouldn't entity level operations like Entity.add_variable and Entity.update_data also invalidate the EntitySet metadata?

[kmax12]

added those and also refactored the metadata reset into a function called EntitySet.reset_metadata()