`testing` improvements

[plajjan]

• [x] rename unit_test_runner to test_runner (and vice versa since there is already a test_runner, which actually just runs unit_tests)
• [x] remove testing.AssertError and replace its use with AssertionError (builtin)
• [x] send Logger to test function
• [x] add ActorTests and EnvTests classes + add to args in `test_runner(env, unit_tests, actor_tests, env_test)
• [x] actually implement SyncActorTests
• [x] actually implement AsyncActorTests
• [x] actually implement EnvTests
• [x] #1631
• [x] implement __test_main discovery in compiler -> produce .test_root.c
• [x] add builder support for .test_root.c
• [x] depends on #1427
• [x] add command to just list all tests
• [x] add option to run a single test
• [x] separate printing status from keeping results up to date etc
  • printing should be a separate actor that receives status updates
  • the current test runner should instantiate such an actor and send all the updates to it
• [x] add support for JSON output
  • building on the previous item of separating printing
  • now becomes possible to run multiple test programs that all output
• [x] do multiple test loops (per default) to verify stability and non-flakiness
  • like 10 loops? or maybe try to run for 0.1 seconds or something?
• [x] support progress reporting, not just final status
  • useful for multiple iterations in test loop
• [ ] build special logger that can capture output and display on error?
• [ ] maybe add support for golden tests?
• [ ] perf timing history
  • [x] record snapshot to a file on disk acton test --record
  • [ ] perf_data file should include history of perf snapshots, not just latest one, this way we can draw a graph over time
  • [x] acton test should read latest snapshot and display perf results including comparison
  • [ ] time module to have humanize() function to display time in human terms, like 123ns or 123.4ms, so pick suitable prefix and format so we get like 4 characters or something compact like that
  • [ ] 12ms +1 14ms -1ms 16ms -1ms