Add support for multi-layer perceptron estimators

[iamDecode]

This PR adds initial support for neural network models. Non-fully connected networks are supported using fully connected multi-layer perceptrons, that simply have a bunch of zero-weight neuron connections for the ones not described in the PMML document.

There is no standard way of dealing with categorical variables in neural networks, hence I have left categorical support out of this PR for brevity. It is currently up to the user to encode categorical variables first.

[codecov]

Codecov Report

Merging #40 (3c5ccc2) into master (45d4b2c) will increase coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #40      +/-   ##
==========================================
+ Coverage   99.62%   99.65%   +0.02%     
==========================================
  Files          19       22       +3     
  Lines        1062     1146      +84     
==========================================
+ Hits         1058     1142      +84     
  Misses          4        4              

Impacted Files	Coverage Δ
sklearn_pmml_model/neural_network/__init__.py	100.00% <100.00%> (ø)
sklearn_pmml_model/neural_network/_base.py	100.00% <100.00%> (ø)
sklearn_pmml_model/neural_network/_classes.py	100.00% <100.00%> (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 45d4b2c...3c5ccc2. Read the comment docs.

[lgtm-com]

This pull request introduces 2 alerts when merging f7fb5d0 into 45d4b2c - view on LGTM.com

new alerts:

• 2 for Conflicting attributes in base classes