Update elastic-package test command to define each command separately

[mrodm]

Relates #787

This PR defines each subcommand under elastic-package test in its own function. Each test runner command now defines:

• the minimum flags for them.
• the needed processing depending on
  • whether or not the type of test can run per data stream.
  • whether or not test folders are required.
  • the package being an input or integration package.
• the minimum fields in TestOptions struct.

This allows us to remove functions from the TestRunner interface: TestFolderRequired, CanRunPerDataStream, CanRunSetupTeardownIndependent

Take the opportunity in this PR to add the test type into the filename written for the xUnit results. This will help us to identify easily which XML file to select/open.

# before
build/test-results/
├── nginx_1717518654614473334.xml
├── nginx_1717518811869606562.xml
├── nginx_1717518815623075388.xml
└── nginx_1717518817400059691.xml

# after
build/test-results/
├── nginx_asset_1717517891228616986.xml
├── nginx_pipeline_1717517893039231373.xml
├── nginx_policy_1717517893079633874.xml
├── nginx_static_1717517893189605328.xml
└── nginx_system_1717518055889708880.xml

The current regex used in Junit step (Buildkite) keeps matching:

elastic-package/.buildkite/pipeline.yml

Line 75 in 1765efb

artifacts: "build/test-results/*.xml"

How to test

elastic-package stack up -v -d
cd /path/to/package

# test each test runner separately
elastic-package test asset -v
elastic-package test static -v
elastic-package test system -v
elastic-package test policy -v
elastic-package test pipeline -v

# test running all tests at once
elastic-package test -v

[elasticmachine]

💚 Build Succeeded

• Buildkite Build
• Commit: 59a9975

History

• 💚 Build #3306 succeeded 2800992
• 💔 Build #3304 failed 53c24d5

cc @mrodm

[jsoriano]

👍 🔥

[jsoriano]

As you prefer, I would be ok with removing this flag from other test runners if it is doing nothing there.

[jsoriano]

Now that we have specific commands for each test runner, maybe we can call here the specific runner, with an specific set of options? This would also allow to remove the registry of tests.

Actually now we wouldn't even need the test runner interface, right? 🤔

[mrodm]

It looks like this could be possible, I'll check this out in a follow-up PR.