Prevent StackOverflowError in ResolverFully

[thc202]

Description

Add guard to resolve schemas just once to prevent loops/stack overflows.

ResolverFully.resolveSchema() threw java.lang.StackOverflowError when parsing
OpenAPI documents that contain recursive schema definitions. For example: a schema
whose additionalProperties.oneOf references itself, or a tree-shaped model whose
items points back to the parent type.

Two-phase guard:

Phase	Mechanism	Keyed by	Catches
In-progress	schemasInProgress	object identity	cycles found during descent
Completed	resolvedSchemas	object identity	repeated resolution of already-resolved schemas

Fix #1751.

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)

[ewaostrowska]

Hi @thc202,
Thank you for providing this PR and working on this long standing issue :)

The solution you came up with has covered most of the cases here, but not all. resolvedSchema only did memorize the already processed schemas, but did not cover the schemas currently being resolved. So the solution would not work for case where: resolving schema A reaches A again through properties, etc before the first call completes or resolving A, the resolver may enter B, which re-enters A before A is in the completed cache (A -> B -> A)

I have introduced in-progress cache as well for that reason.

[thc202]

Thank you!