Added groups param to convolutions

[Michaeljurado42]

Motivation and context:
There is a (tentatively) accepted pull request for adding grouped convolutions to nengo already. Before adding grouped convolutions to nengo-loihi it is neccessary to add grouped convolutions to nengo-dl.

Related Forum Posts
https://forum.nengo.ai/t/adding-a-groups-parameter-for-convolutions-in-nengo-loihi/2035/2
https://forum.nengo.ai/t/trying-to-understand-nengo-transforms-convolution/1844/4?u=mjurado3

Interactions with other PRs:
This change is dependent upon this pull requrest in the nengo repository: nengo/nengo#1684

How has this been tested?
This is tested through the test_conv in nengo_dl/tests/test_converter.py

How long should this take to review?
30 minutes if there are no additional features that should be added.

Where should a reviewer start?

1. First the reviewer should look at ConvIncBuilder and convince themselves that no change is needed in this function for adding grouped convolutions. I was somewhat surprised when the unit tests passed but then I realized that tf.nn.convolution infers the groups parameter from the shape of the weights as shown in this example
2. Then the reviewer should verify that test_conv in nengo_dl/tests/test_converter.py provides sufficient coverage.

Types of changes:
New-Feature: High Level addition of groups parameter to ConvertConv

Checklist:

• I have read the CONTRIBUTING.rst document.
• I have updated the documentation accordingly.
• I have included a changelog entry.
• I have added tests to cover my changes.
• I have run the test suite locally and all tests passed.

[tbekolay]

Should have a period at the end, and link to the pull request. See the other entries in this file for examples.

[tbekolay]

Note to reviewer: when testing things as @Michaeljurado24 has described in the PR description, make sure these things work in the older versions of TF that we support.

[drasmuss]

Merged, thank you for your contribution!