Add trace output to the test mode

[mlbiam]

When running opa test I don't get any trace output, just the result. To get the trace output I need to run the opa run command line and enable trace. If this feature were available for opa test it would make for a much quicker development cycle

[tsandall]

Let's extend opa test -v to print the trace for failed tests.

For example, without -v OPA only prints failures:

$ opa test .
data.tests.test_not_equal: FAIL (460ns)
--------------------------------------------------------------------------------
PASS: 4/5
FAIL: 1/5

With -v OPA prints all test names:

$ opa test . -v
data.tests.test_single: PASS (565ns)
data.tests.test_multiple: PASS (740ns)
data.tests.test_mismatch_one: PASS (558ns)
data.tests.test_not_equal: FAIL (555ns)
data.tests.test_gt: PASS (647ns)
--------------------------------------------------------------------------------
PASS: 4/5
FAIL: 1/5

Let's extend -v to also print traces for test failures:

FAILURES
--------------------------------------------------------------------------------
data.tests.test_not_equal (FAIL) (555ns)

   <trace output goes here>

SUMMARY
--------------------------------------------------------------------------------
data.tests.test_single: PASS (565ns)
data.tests.test_multiple: PASS (740ns)
data.tests.test_mismatch_one: PASS (558ns)
data.tests.test_not_equal: FAIL (555ns)
data.tests.test_gt: PASS (647ns)
--------------------------------------------------------------------------------
PASS: 4/5
FAIL: 1/5

[tsandall]

See my previous comment for details on expected behaviour.

The required changes can be made in:

• cmd/test.go: if verbose is true, instantiate topdown.BufferTracer and set on the test runner.
• tester/reporter.go: update JSONReporter and PrettyReporter to emit trace results.

[srenatus]

🎉 I think is is done with #978 and #975, isn't it? 😃

[tsandall]

Fixed by #975 thanks to @srenatus!