Summary:
run_tests_on_spark.py script has three (related) improvements:
* When dynamically generating test list from C++ binaries, just collect them locally without a
  spark job. This avoids the spark overhead and can be done in a short time (~90 second) that would
  normally be idle waiting for spark resources.
* A new ignore_list option that can be used to remove tests from the generated test list. This can
  be used when re-running a test job that did not complete and the missing tests are unknown. By
  passing the list of tests already run & passed, the new run can complete the job efficently.
* A new disable_list option that can be used to remove tests from the generated test list, and also
  report them as skipped. This can be used to provide a list of disabled (aka muted) tests that need
  not be run.

Test Plan:
Jenkins: build type: fastdebug
Additional jenkins work needed to fully exercise new features.

Reviewers: jharveysmith

Reviewed By: jharveysmith

Subscribers: devops

Differential Revision: https://phorge.dev.yugabyte.com/D46553