Address pain points from latest onboarding

[jcharum]

We had a few recent new Bigslice users. This is an umbrella issue that covers several small things we're going to do to try to address some of the pain and confusion they experienced.

• Make the requirement for fat binaries when we have driver-worker architecture mismatch more prominent.
• Add OS X to our bigslice and bigmachine CI (just to make sure if folks' first touch is on OS X local execution, it all will work). Expand matrix with Go 1.14 and OS X #98 for Bigslice; Add GitHub actions CI bigmachine#50 for Bigmachine, as part of migration to GitHub Actions.
• Make clear that we only currently support x86/Linux workers. Add footnote for linux/amd64 cluster nodes #106
• Add documentation of usage of Err() to Scan(), so that it’s harder to miss. Suggest calling Err in Scan(v)? documentation #101
• Create a codewalk that makes this (and other parts of a starter program) more clear upon first encounter.
• Add an example test of a simple user-created slice operation. Add example of slice testing #109