add code without coverage to verify that a decrease in coverage fails the build

[m-mcgowan]

No description provided.

[m-mcgowan]

As well as this PR, the settings in coveralls.io were changed:

• coverage threshold for failure: 80%
• coverage decrease threshold for failure: 0.1%

[m-mcgowan]

The coverage difference is computed against the last two commits to this branch, which is not what we want.

The job ID seems a bit strange, which is indicated here as the cause - github.com/lemurheavy/coveralls-public/issues/1019. (Deliberately not putting the full url so we don't create a reference to this PR from that issue.)

Going to try to set the job ID from the Github CI environment.

[m-mcgowan]

I like the github action better than our previous approach because it gets branch names and other details correct, although it doesn't appear check the PR against the master branch at present. (That may have something to do with the type of reporting used, so we'll check after merging this to master.)

[bsatrom]

If this is ready for review, I can give it a look.

[m-mcgowan]

I think it's about as good as we can get it considering the uphill climb it's been to get this far. I'd like to merge it and gain some experience with how well it works and then make adjustments as necessary.

[zfields]

I'll review it today.

[zfields]

I spent the better part of the day reviewing the PR and understanding the short-comings of the official Coveralls GitHub Action...

The official Coveralls GitHub Action does not distinguish between the Docker action file system, VM action file system and native GitHub runner file system.

By updating the file paths in the lcov file, I was able to eliminate a significant portion of our CI infrastructure and simplify the implementation. That research spike is captured in #67.

I propose closing this PR and adopting #67 instead.

[zfields]

Propose migrating to alternate solution.

[zfields]

Closed in favor of #67