Add recipe specification

[therazix]

This PR introduces a draft of a recipe specification. The recipe can be used by tmt to reproduce a run with the exact same configuration, without requiring additional input. The recipe file contains preprocessed data about tests, results, and run configuration to eliminate the need for dynamic evaluation.

Some notes:

• Recipe contains tests with all adjustments already applied. Since context variables and environment settings have been evaluated and propagated to individual tests before recipe generation, do we need to preserve the original environment configurations from different levels (command-line, plan)?
• How should test requirements be handled? Tests may require additional files. Where should we save them? A similar issue is with Ansible playbooks, but these could potentially be stored directly inside the recipe.

Resolves: #3854
Related: TT-273

Pull Request Checklist

• implement the feature
• write the documentation
• extend the test coverage
• update the specification
• adjust plugin docstring
• modify the json schema
• mention the version
• include a release note

[psss]

Thanks for kicking this off! Added a couple comments.

[psss]

Perhaps it would be good to also mention here the use case which was the main motivation for creating the concept of recipes: Processing the recipe and filtering tests in order to submit only a subset of the discovered tests.

[psss]

In order to be able to reproduce the run we need the git repo details as well. It should be basically fmf id, just without the name, so url of the repo, ref in the form of a hash and path to the metadata tree. The name part should be covered by plan names under the plans key below.

Also, I guess, we should store here the original environment and context with which the run was created? The complete environment should be stored for each test in tests.yaml, but the environment variables are also delivered to other places, such as scripts executed during the prepare step.

[psss]

Shall we explicitly mention that when submitting a new run based on a recipe the results key is ignored and new results will be created based on the actual test execution? Or do we expect some integration of the past results as well?

[psss]

In order to be able to execute tests the expanded test metadata will not be enough. Although we do not expect any adjust rules to be evaluated or any other dynamic steps, we will have to fetch the git repositories again so that the test code itself is available. Perhaps a short mention of this should be added here or in the story description?