100% coverage

[hunse]

Get 100% coverage in codecov.

• [x] Cover all BuildErrors and other exceptions
• [x] Remove unused code
• [x] Test code that should not fail in tests (e.g. chip connection/communication) using mocks
• [x] Cover any remaining uncovered lines, or mark as "no cover" (e.g. failed imports)

[hunse]

OK, I think I've got everything covered, except for the few cases that we've explicitly talked about.

Currently, this is still a lot of commits. I've organized them into added tests first, then code removals, then other stuff. I've kept them separate for now (I'm thinking it will be easier to review that way), but they can mostly be squashed before merging (maybe squashing all the tests into one commit, and most of the removals into another). The two removals we should keep separate IMO are the first two in the order right now: e29050a and dfa4da5. The first removes the NIF neurons and replaces them with SpikingRectifiedLinear for our node encoding. The second gets rid of probes from LoihiInput, which we don't use right now, but may want to add back in in the future to be able to probe parts of host->chip connections.

I also think we should keep the "other stuff" mostly separate (i.e. things that are not tests or removals), since they're all fairly different changes and some have potential functional consequences.

This is also based off #140 right now (to fix the tests), so we should merge that first.

[hunse]

One thing we haven't discussed is conftest.py. It technically has one uncovered line. I would recommend removing it from coverage entirely.

[hunse]

I've added changelog entries. There were only four commits that I thought deserved them.