Investigate how many APIs can be tested in parallel by nimibot.

[ni-jfitzger]

Description of issue

There are 3 nimibot runners.

• 1 for Linux
• 1 for 32-bit Python on Windows
• 1 for 64-bit Python on Windows

By default, if max-parallel is not specified in the workflow, a value will be used based upon runner availability. Since we seem to only test 3 APIs on each nimibot at once, I suspect this is being set to 3 based upon the existence of 3 runners, despite the fact that each workflow is restricted to a specific runner. We might be able to set it higher and test more APIs at once, reducing the time that it takes to complete testing.

(We should keep in mind that there are shared resources, like the grpc_device_server and some multi-threading tests). I also don't know how many processors these VMs have available. It may or may not be a good idea to set max-parallel all the way to 9.

[ni-jfitzger]

Since we seem to only test 3 APIs on each nimibot at once, I suspect this is being set to 3 based upon the existence of 3 runners, despite the fact that each workflow is restricted to a specific runner.

I don't think that's what's happening, after all. For a run on master, we had 5 APIs being tested at once.