fix: AnnotationUtils cache misses due to reinstantiation of ModelConverterContext

[ldt-fweichert]

Description

ModelConverterContextImpl stores each already processed type and it's corresponding Schema internally. When types are discovered multiple times during resolution, they should be converted only once. Multiple usages of static helper functions on AnnotationUtils instantiated a new ModelConverterContext instead of reusing the initial one effectivley ignoring the cache of processed types.

This PR adds a ModelConverterContext context-Parameter to some AnnotationUtils helpers to reuse the instance during resolution. This speeds up resolution, especially with Recursive Models, where resolving Models with 10 subtypes could require at least multiple hours of processing time.

Two questions from my side remain:

• Should ModelConverterContext be instantiated inside AnnotationUtils at all? If no, the overloads created in this PR should be removed entirely, to ensure correct usage. An additional null-check should be added as well.
• My added test of checking for big differences in timing between models with 2 subtypes and models with 10 subtypes is admittedly not that maintainable. I'd be open to alternative ideas for testing, although it served my use-case pretty well.

Fixes: #5091

Type of Change

• 🐛 Bug fix
• ✨ New feature
• ♻️ Refactor (non-breaking change)
• 🧪 Tests
• 📝 Documentation
• 🧹 Chore (build or tooling)

Checklist

• I have added/updated tests as needed
• I have added/updated documentation where applicable
• The PR title is descriptive
• The code builds and passes tests locally
• I have linked related issues (if any)