Add some missing layers commonly used for image processing

[Octogonapus]

Description of the Change

This PR adds the AveragePooling2D, GlobalMaxPooling2D, SpatialDropout2D, UpSampling2D, and GlobalAveragePooling2D layers. This PR also fixes a logging problem with the kotlintest artifact and adds a new layer type, Layer.ModelLayer, for elementary TF 1.15 support.

Motivation

These layers are commonly used for image processing, which is our initial use case. Along the way, I also had to fix some problems with logging and TF 1.15 (see #98).

Verification Process

New tests. Tested some generated code manually. Added a new test that loads a model for transfer learning with mobilenet on TF 1.15.

Applicable Issues

None.

[codecov-io]

Codecov Report

Merging #99 into master will increase coverage by 2.15%.
The diff coverage is 93.75%.

@@             Coverage Diff              @@
##             master      #99      +/-   ##
============================================
+ Coverage     89.67%   91.81%   +2.15%     
  Complexity        1        1              
============================================
  Files            76       77       +1     
  Lines          1587     1698     +111     
  Branches        175      187      +12     
============================================
+ Hits           1423     1559     +136     
+ Misses          100       71      -29     
- Partials         64       68       +4