Update RNG

[alanlujan91]

Since Numpy version 1.17.0 scipy introduced Generator as a replacement to RandomState.

This PR updates HARK to use Generator as recommended in https://numpy.org/doc/stable/reference/random/index.html#quick-start

• Tests for new functionality/models or Tests to reproduce the bug-fix in code.
• Updated documentation of features that add new functionality.
• Update CHANGELOG.md with major/minor changes.

[alanlujan91]

Few errors remain regarding precision given a new RNG.

[llorracc]

Are these the kinds of errors that would disappear if we scaled back our degree of precision? If so, that's the solution.

[alanlujan91]

TODO: level of precision to 3 or 4 decimals global default

[alanlujan91]

FIND: assertAlmostEqual((.), (\d.\d{5}))
REPLACE: assertAlmostEqual($1, $2, places = HARK_PRECISION)

[llorracc]

@alanlujan91,

Thanks for this! Some people dream of winning or beauty or fame. It has long been my dream to reduce the excessive precision of our numerical tests. This is a big step in that direction!

Comments:

1. I think we might want at least a couple of different precision defaults
   • HARK_PRECISION_NORMALIZED for variables that are normalized by permanent income
   • HARK_PRECISION_ZERO for variables that should be approximately zero
2. Sadly, your PR ruins some beautiful formatting in which explanatory comments were perfectly aligned with each other on the right hand border of some multiline code snippets. Now everything is mooshed to the left to be one space beyond the last character of the actual instruction.
   • This is probably the fault of VSCode or Black or Flake8. It's not worth a lot of effort to fix -- but is worth a little. So if it only takes a little, please do it.

[codecov-commenter]

Codecov Report

Base: 73.62% // Head: 73.61% // Decreases project coverage by -0.01% ⚠️

Coverage data is based on head (1506245) compared to base (86d5c62).
Patch coverage: 75.29% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1193      +/-   ##
==========================================
- Coverage   73.62%   73.61%   -0.02%     
==========================================
  Files          72       73       +1     
  Lines       11973    11998      +25     
==========================================
+ Hits         8815     8832      +17     
- Misses       3158     3166       +8     

Impacted Files	Coverage Δ
HARK/ConsumptionSaving/ConsAggShockModel.py	89.64% <ø> (ø)
HARK/ConsumptionSaving/ConsGenIncProcessModel.py	82.45% <ø> (ø)
HARK/ConsumptionSaving/ConsPrefShockModel.py	79.68% <ø> (ø)
HARK/ConsumptionSaving/ConsRepAgentModel.py	100.00% <ø> (ø)
HARK/ConsumptionSaving/ConsRiskyAssetModel.py	40.59% <ø> (ø)
HARK/estimation.py	0.00% <0.00%> (ø)
HARK/utilities.py	30.09% <22.80%> (+0.46%)	⬆️
...RK/ConsumptionSaving/tests/test_ConsMarkovModel.py	86.56% <60.00%> (ø)
HARK/interpolation.py	43.34% <60.00%> (ø)
HARK/ConsumptionSaving/ConsMarkovModel.py	85.83% <66.66%> (ø)
... and 19 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[llorracc]

@alanlujan91 let's discuss merging this in tomorrow's meeting.