Add samediff layer support (forward only, no output layers)

[AlexDBlack]

Starting from: #4426 - removes the backprop components and output/loss layers... I'll tackle those components separately.

Adds support for SameDiff layers:

• Forward pass only for now, non-output layers only. Backprop is partially implemented, but disabled for now
• I've tried to keep everything to a single class only: that means, implementing a SameDiff layer requires only extending BaseSameDiffLayer, and that's it (no implementation or param initializer classes, for example).
  • 2 complete examples of samediff layers can be found in this PR (both in test directory) - SameDiffConv and SameDiffDense. Param initialization and forward pass match the DL4J layers.
  • See also MinimalSameDiffDense for a minimal version (least number of methods required to be implemented)

[AlexDBlack]

Something we will want to support: importing a layer definition (serialized SameDiff, TF, etc) into a DL4J layer. The current layer API may not easily support this in all cases.

[wmeddie]

Looks pretty good.

One thing I'l like to see if to minimize the number of functions you need to override.

One way we can do that is to define helper methods in BaseSameDiffLayer that you could use when creating the graph. These methods can save the parameter attributes and for calculating the values for weightKeys, biasKeys, paramShapes and initializeParams.

For example, we might be able to define an optional constructor (or overridable method) that takes a pair of arrays of SDVariables for weights and biases and use their sizes to calculate the shapes. The Chainer API takes that approach.

I'm not familiar with when the keys and shapes are required though, so it may not be possible to implement SameDiffLayer that way.

[wmeddie]

I felt this would be better defined as a List<SDVariable> instead of List<String>. That way you aren't just throwing away objects, and it's impossible to make a typo.

[wmeddie]

Here's an idea oh how it could work (and a working unit test)

https://gist.github.com/wmeddie/809c602311503a2ebb2d2cd957daf09e