Prevent pollution between scenario tests

[kasey]

In a few e2e evaluators, package-level variables are used to persist data across evaluator calls, which are otherwise stateless. In the scenario test runs, where we execute multiple tests in the same bazel command, these variables are contaminated on all but the first test. I think we should redesign these evaluators with the understanding that some can't be stateless, but all need to reset between runs. This PR moves stateful values from package-level variables in the evaluators package, into the EvaluationContext that is already passed to all evaluators, which is re-initialized between runs.

What type of PR is this?

Bug fix

What does this PR do? Why is it needed?

This fixes a spurious test failure for the TestEndToEnd_MinimalConfig_ValidatorRESTApi target, specifically this evaluator:

    --- FAIL: TestEndToEnd_MinimalConfig_ValidatorRESTApi/voluntary_has_exited_8 (0.00s)