[#315] Respect missing bias

[bzfwunde]

If in a pytorch layer no bias is provided, we errornously used the bias from the previous layer. We must use 0 bias instead. Added extend pytorch example thanks to Pierre.

This addresses #315. @fuadcan, could you please test that it fixes your issue? Thanks!

[CLAassistant]

All committers have signed the CLA.

[pobonomo]

I'll probably have forgotten by next time we need to do it but here is how I generated the new test network for pytorch (adding one layer without bias): https://gist.github.com/pobonomo/8d429bdc381690328868e684867ff972