Scorecard2 pod execution

[jmccormick2001]

Description of the change:
this PR adds functionality to the alpha scorecard implementation, you can test
this PR out by running the following command as an example:

operator-sdk alpha scorecard \
--config=internal/scorecard/alpha/testdata/config.yaml \
--selector=suite=basic \
-o json \
--bundle=internal/scorecard/alpha/testdata/bundle/ \
--cleanup=true

this PR includes the following:

• creates a ConfigMap that holds the on-disk bundle contents, this ConfigMap is passed to test pods
• test pods are created by the scorecard for selected tests
• test pod logs are fetched and converted into Scorecard test output for viewing results
• output is formatted into json or text as the case with the current scorecard
• a cleanup flag is optional if you want to have access to the test pods for debugging

Motivation for the change:

this is part of a larger set of work to implement the alpha scorecard, but this PR is mostly focused on wiring up the scorecard command to have test pods created, and test pod log transformed into scorecard output format for user reporting.

[camilamacedo86]

What is the bundle zip ?

[camilamacedo86]

Could we add a comment explaining ?

[jmccormick2001]

good catch, originally I was using zip and then switched to tar to archive up the bundle contents, I've updated this and added a comment.

[joelanford]

Rather than logging an error, we should convert this error to a ScorecardTestResult and append it to the list of results. Ditto for any other errors related to anything happening in a test run.

[jmccormick2001]

fixed.

[joelanford]

This implementation is going to make it difficult to write unit tests for the RunTests function. WDYT about making a TestRunner interface to use instead.

type TestRunner
   func RunTest(context.Context, test Test, bundle *registry.Bundle) (*TestResult, error)
}

And then have the main struct (Options right now, maybe Scorecard based on another comment), have something like:

type Scorecard struct {
	Config     Config
	Selector   labels.Selector		
	WaitTime   time.Duration
	Bundle     *registry.Bundle
	TestRunner TestRunner
}

var _ TestRunner = &PodTestBundleRunner{}

type PodTestRunner struct {
	Client         kubernetes.Interface
	Namespace      string
	ServiceAccount string
	SkipCleanup    bool
}

func (r *PodTestRunner) RunTest(ctx context.Context, test Test, bundle *registry.Bundle) (*TestResult, error) {
	return actuallyRunTheTest(ctx, test, bundle)
}

var _ TestRunner = &FakeTestRunner{}

type FakeTestRunner struct {
	TestResult *TestResult
	Error  error
}

func (r *FakeTestRunner) RunTest(ctx context.Context, test Test, bundle *registry.Bundle) (*TestResult, error) {
	return r.TestResult, r.Error
}

Then on the cmd CLI side, we can construct the TestRunner and set that field on the scorecard struct.

[jmccormick2001]

interesting approach, but I'm wondering if we could address this in another PR since I don't think it impacts the test interface and I'd like to unblock the folks writing tests currently. I think there is a story to rewrite the scorecard tests, maybe we could do this in that story?

[joelanford]

LGTM