Add support for Richer Input in OONI Run v2 backend

[hellais]

We would like to add support for returning richer input specifications inside of OONI Run v2 descriptors.

This has the nice benefit that we would not have to pass through a dedicated API endpoint and we can potentially also make use of the caching which happens by default in in-app descriptor loading.

As part of this epic we should:

• Collect some of the requirements for richer input (started in the engine simplification doc)
• Come up with a new revision of the spec
• Implement it in the backend
• Implement support for it in the engine