Improve ADT validators' performance

[kubukoz]

Closes #1319. Also actually prevents ADTs from being mixins, which shouldn't have been possible in the first place.

Tested on a model with 92k shapes (including 78k members) using 288 @adtMembers and 33 @adts.

Before: 11+ seconds

The original implementation was spending most its time in the getReferences method:

which had linear time complexity on the number of member shapes in the model, which let's be honest - are the most common type of shape by far in any realistic model.

After: <200ms

In the new implementation, there's no sharing of code (the validation patterns are different, so I didn't think it's worth sharing), and we utilize things like getShapesWithTrait and selectors to get better perf. Probably linear in the number of shapes annotated with each of the ADT traits.

PR Checklist (not all items are relevant to all PRs)

• Added unit-tests (for runtime code)
• Added bootstrapped code + smoke tests (when the rendering logic is modified)
• Added build-plugins integration tests (when reflection loading is required at codegen-time)
• Added alloy compliance tests (when simpleRestJson protocol behaviour is expanded/updated)
• Updated dynamic module to match generated-code behaviour
• Added documentation
• Updated changelog

[kubukoz]

So far, so good - AdtTraitValidator is all but gone from the flame graph.

Baseline: #1319 (comment)

[kubukoz]

Got the AdtMemberTraitValidator too.