Implement a 1d convolution followed by a batch norm

[julianhoever]

Implement a batch normed 1d convolution similar to the batch normed linear layer (see #240)