Evaluate alternative test suite structure

[niknetniko]

The current JSON Schema and DSL parsing code is unnecessarily complex. Part of this might be fixed with ideas from #586.

Another part is due to the fact that we support an alternative naming scheme for test suites. These were added in #424 in the context of a paper to make TESTed less linked to Dodona, but I'm not sure there are practical uses for maintaining these. They have never been in the documentation as far as I know, nor been recommended to external users.

If these are actually used and we want to keep supporting this, I would try to extract it into a separate step that is run before the actual DSL.

So:

• Are there exercises that use this naming?
• Do we want to keep supporting this?
• Do we need a compatibility layer if we decide to remove it?
• If we decide to keep it, how can we improve maintainability of the JSON Schema and other DSL-related code?

[pdawyndt]

I have adopted the new naming scheme.

[niknetniko]

I see, good to know. In that case we'll probably shift to finding ways to improve maintainability of the JSON Schema and parser.

[pdawyndt]

@niknetniko I would also be OK if we could transform all occurrences of the new naming scheme into the old one. All have been generated, so this would also require to update the generators in case we need to regenerate test suites.

[bmesuere]

The 2 alternative schemes also make it hard to provide a useful autocomplete since there are twice as many keys.