Fix shape equality in ADT validator

[kubukoz]

It was causing false negatives (doesMemberTargetAdtShape being false when it should be true) for some shapes, when a lot of shapes were added in an unrelated namespace.

Some ShapeId constructors use a cache internally, which explains why it'd sometimes still match instance equality, but it's not a mechanism that should be relied upon, and it will never be one.

I've also checked for any other usages of == in our Java sources and they are all about primitives/enums, so it's fine.

Not adding tests due to the unpredictable nature of the issue.

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]

update:

the ShapeId factory/cache has a size bound, so I'm pretty sure the users who bumped into this were simply adding enough new shapes to go over this limit before the @adt-marked shapes had their ShapeIds created - causing the latter to never get cached, and thus never have referential equality match.