Add convolution operation

[vascoorey]

Adds support for convolving matrices with NxM kernels (not necessarily square kernels) using the implementation described in https://developer.apple.com/documentation/accelerate/vdsp/2d_convolution