[experiment] Run Comet tests in Docker

[andygrove]

Which issue does this PR close?

Part of #1786

Rationale for this change

Exploring the idea of running the tests in Docker so that we can run them locally exactly the same way that they run in CI.

What changes are included in this PR?

How are these changes tested?

[andygrove]

This approach seems to work, and the Spark 4 tests ran a little faster than usual. Total job time was 87 minutes.

Run completed in 1 hour, 10 minutes, 53 seconds.
Total number of tests run: 1044
Suites: completed 37, aborted 0
Tests: succeeded 1044, failed 0, canceled 8, ignored 58, pending 0
All tests passed.