Reconsider matching Enceladus runs

Enceladus runs may change state when multiple attempts are made. The failed attemts may or may not result in state changes. This complicates matching individual spark attempts to appropriate runs.
The current way relies on order of attempts and runs, which is not sufficient. (e.g. sometimes, when two attempts fail, the first attempt manages to update run doc while the second does not.)

For a quick solution one may compare the run's startDateTime (for Std) and start of Conformance checkpoint (for Conform) to be within the attempt duration.

In a long term, the Enceladus api and metadata should be extended to include attempt id.