Fix batch handling for Add::Gradient()

[rcurtin]

The current implementation of the Add layer (which just adds a bias to each element) does not properly handle batch sizes greater than 1. This can be seen in the example in #3666.

The Gradient() function expects the error matrix to be of size outSize x batchSize, and the gradient matrix itself is already sized to be weightSize x 1. (For the Add layer, it is conveniently true that weightSize == outSize.)

However, the existing implementation of gradient = error is only correct when batchSize = 1. Much like the Linear layer, the bias term's gradient is computed as the sum across all input points in the batch.

I then noticed that the Add layer is not tested, so I added a simple test for it at a batch size of 1 and also at a larger batch size. (Previous to this PR, the second test will fail.)

[mlpack-bot]

Second approval provided automatically after 24 hours. 👍