fix: Handle list[Model], dict[str, Model] and Optional[Model] in nested type detection

The nested schema type detection from PR #426 did not resolve $ref for
list items, dict additionalProperties, or anyOf (Optional) fields.

This caused KeyError or incorrect type inference when using Pydantic or
dataclass models with fields like list[NestedModel], dict[str, NestedModel],
or Optional[NestedModel].

Changes:
- Add $ref resolution to _get_element_type for centralized handling
- Extract _resolve_ref, _resolve_property_type, _resolve_single_type,
  and _resolve_typed_property helpers to eliminate duplicated $ref logic
- Properly validate Optional via anyOf (check len==2 and null second element)
- Add default=None for optional fields not in required list
- Include optional properties in property_order so fields with defaults
  are not skipped during schema traversal
- Extend __init__ wrapper to convert list[dict] and dict[str, dict]
  to their respective nested types

Ref: flyteorg/flyte#6887
Signed-off-by: André Ahlert <andre@aex.partners>

Author: andreahlert