first implementation of Conv1D from EJ

Related to issue #13 (closed)

Took @ejk43's implementation of Conv1D layer from https://github.com/Xilinx/RFNoC-HLS-NeuralNet/blob/master/rfnoc/hls/nnet_lib/nnet_conv.h

Things changed:

• removed hls::streams
• changed output logic so output size is the same as input size (zero padding added only to the left of inputs)

Things to update/make configurable:

• configurable "zero" padding: valid, same, full, etc.
• number of filters
• reuse factor

@benjaminkreis also has some ideas about trying to make everything more parallelizable