Validate test cases against a schema

[echeran]

This PR is to add automation so that we guarantee that our test data remains in a consistent state. This prevents problems we had previous of ill-formed JSON and/or invalid contents according to the schema.

This PR moves the work from mradbourne#1 to the upstream, given that #767 was merged already without this addition in order to keep things simple for #767.

From the discussion there, we have a couple of options for how we go about automating the test validation:

• Option 1: write a custom script to validate test cases' JSON files
  • Pros: Can add logic to pick up the latest schema version
  • Cons: We might have to maintain code in the future
• Option 2: use an existing tool (ex: ajv-cli) to validate test cases' JSON files
  • Pros: Use an off the shelf tool in JS for JSON
  • Cons: We have to remember to maintain the CI workflow to when there is a new schema version

Aside: A decision point from #767 was the format of the version number for the schemas - should it be semantic versioning or an auto-increment integer or something else. #767 changed to auto-increment integer, which I have also have a slight preference for. We're all in agreement there, AFAICT.

[eemeli]

We do not need a custom script for this. My suggestion in mradbourne#1 (comment) works to replace the current behaviour, but without a custom Python script.

If we want to automatically use the latest schema version, here's how to do that:

ajv validate --spec=draft2020 \
-s $(ls -1v schemas/*/*schema.json | tail -1) \
-d 'tests/**/*.json'

[echeran]

If we want to automatically use the latest schema version, here's how to do that:

ajv validate --spec=draft2020 \
-s $(ls -1v schemas/*/*schema.json | tail -1) \
-d 'tests/**/*.json'

This updated command handles the question of getting the latest schema version. I'm all for not having to write & maintain something if we don't need to, so let's go with that.

Just to make sure that it correctly sorts numerically and doesn't fallback to lexicographic sorting, I did a mini test:

$ cd /tmp
$ mkdir -p schemas/v{1,2,3,4,5,6,7,8,9,10,11}.0.1
$ touch schemas/v{1,2,3,4,5,6,7,8,9,10,11}.0.1/tests.schema.json
$ mkdir -p schemas/v11.1.0
$ touch schemas/v11.1.0/tests.schema.json

$ ls -1v schemas/*/*schema.json 
schemas/v1.0.1/tests.schema.json
schemas/v2.0.1/tests.schema.json
schemas/v3.0.1/tests.schema.json
schemas/v4.0.1/tests.schema.json
schemas/v5.0.1/tests.schema.json
schemas/v6.0.1/tests.schema.json
schemas/v7.0.1/tests.schema.json
schemas/v8.0.1/tests.schema.json
schemas/v9.0.1/tests.schema.json
schemas/v10.0.1/tests.schema.json
schemas/v11.0.1/tests.schema.json
schemas/v11.1.0/tests.schema.json

I used semantic versions instead of the autoincrement integer versions that we are currently using just to cover all bases.