Test stability and recovery improvements

[pwrightcertinia]

fixes #67, fixes apex-dev-tools/tooling-issues#92, fixes apex-dev-tools/tooling-issues#84, fixes apex-dev-tools/tooling-issues#52

The main goal with this is to help reduce the number of fatal crashes that occur during test runs. It does not care what errors come in yet (unlike polling client, which only skips network errors), will keep polling or retrying until either hitting a timeout or a max number of tries.

• Integrated the library underneath PollingClient which is ts-retry-promise.
• Polling no longer exits on unknown errors, it will continue until timeout.
• Test runner will return the latest results, the extra dangerous query (gatherResults) after the last successful poll is no longer required.
• Because of the above, it is now also possible to return and generate partial results as long as at least one poll has worked.
• Added retries in various places:
  • Async/sync test start requests
  • Every query call
  • Added a run func to the query helper which will retry any connection based call
  • Abort exec anon
• Retries will default to 3 extra attempts, with 15/30/60 seconds delays - not configurable yet
• Removed test cancel polling - not sure this worked as intended, it did no corrective action and was checking records it supposedly just updated. So this is now fire and forget if the anon apex succeeded.
• Added a lot of additional logging:
  • Every poll and retry that fails will log the reason
  • Waiting for a retry
  • Test failures will print as they come in on status polls
  • Report generation success/failure