Refactor FlexibleType to use TypeBounds instead of case class

[HarrisL2]

Refactors FlexibleTypes to have TypeBounds as underlying representation, which can help with macro-related code avoid match errors.

The key change is in Definitions.scala

How much have you relied on LLM-based tools in this contribution?

Minimally, for some refactoring

How was the solution tested?

Covered by existing tests (this is a refactoring)

[sjrd]

I don't think this is going in the right direction. Superficially, this may allow some macro test suites to pass in the short term. But the logic will probably be broken anyway. And in the long term, this will just force macros to handle flexible type specially anyway, like the compiler. IMO we should stick with the correct design with real FlexibleType, and adjust the few macros that have chosen to pattern match on type shapes. That is fragile, and will eventually fail for other reasons.