JEDI-CI performance fixes

[eap]

Performance improvements for jedi-ci

1. Part of the test-launcher container is pre-built and pushed to the github container repo. This pre-building improves the speed of the action init. We do not pre-build the full image because it would prevent testing of unsubmitted changes.

2. Use a thread pool to cancel batch jobs and GitHub check runs. The functions used to to cancel prior tests runs are network bound and somewhat slow. Rather than running them in serial, this change runs them in parallel with a thread pool executor (Python's GIL doesn't matter here because the latency comes from non-blocking network calls).

Why was this needed?

The JEDI-CI launcher is somewhat interactive and a failure here causes a great deal of downstream developer pain. I want this action to execute as fast as possible in order to improve the chances that a failure is noticed before the user "walks away". Additionally, this will be JCSDA's most widely used GitHub-hosted action and this will prevent it from consuming unnecessary compute time (which is "expensive" due to our limited GitHub-earmarked funding).

Checklist

• I have performed a self-review of my own code
• N/A I have made corresponding changes to the documentation
• I have run the unit tests before creating the PR

[ashley314]

Looks good, thanks!