Add ModelError as else case to lr_normalizer()

[emogden]

• make sure that all tests are passed - done
• create some unit tests and update testing procedures accordingly - done
• make PR to dev (or daily-dev) - done

After adding Talos to my Tensorflow+Keras model, I found that the learning rate normalizer only recognizes objects from 'keras.optimizers' and was failing silently. Rather than making structural/dependency changes to accommodate Tensorflow, I figured it should at least have an 'else' case with an error to indicate the function cannot operate on the provided optimizer.

Also added test function to make sure this error gets triggered with a string.

My first PR, so let me know if this is prepared incorrectly.

[pep8speaks]

Hello @emogden! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! :beers:

Comment last updated at 2019-08-26 20:41:50 UTC

[coveralls]

Pull Request Test Coverage Report for Build 471

• 2 of 2 (100.0%) changed or added relevant lines in 1 file are covered.
• 5 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.4%) to 90.072%

---

Files with Coverage Reduction	New Missed Lines	%
talos/reducers/permutation_filter.py	5	66.67%
<!--	Total:	5		-->

Totals
Change from base Build 450:	-0.4%
Covered Lines:	1125
Relevant Lines:	1249

---

💛 - Coveralls

[mikkokotila]

Nice, thanks :) Later this week there will be the final changes to v.0.6.2 and given this is already passing, I don't see why it would not pass then. Will merge at that point.

thanks again!

[mikkokotila]

Thanks for this :) The issue has been handled now in other commits, so closing here.