Describe the bug:

PatchDict stops honoring Pydantic Field constraints declared on the wrapped schema. Any validators such as max_length, ge, regex, etc. disappear from the generated JSON Schema and are not enforced when parsing patched payloads. This makes partial update inputs accept values that the base schema would correctly reject.

Steps to Reproduce:

1. Define a Ninja Schema that uses Field constraints:

   from ninja import Field, PatchDict, Schema
   from pydantic import BaseModel
   
   class MySchema(Schema):
       foo: str = Field(max_length=3)
   
   MySchemaPatch = PatchDict[MySchema]
   
   class Body(BaseModel):
       my_schema: MySchema
       my_schema_patch: MySchemaPatch

2. Inspect Body.model_json_schema(). my_schema shows maxLength: 3, but my_schema_patch doesn't.

3. Validate a payload with foo="barbar" (length 6) under my_schema and my_schema_patch.

Expected behavior: Both the JSON Schema and runtime validation for my_schema_patch should retain the constraints defined on the base schema, so foo="barbar" should raise a validation error.
Actual behavior: my_schema_patch silently accepts the invalid value. Constraints are absent from the patch schema, so clients can send data that violates the original schema whenever they use partial updates.

Versions:

• Python version: 3.13.1
• Django version: 5.2.8
• Django-Ninja version: 1.5.0
• Pydantic version: 2.12.4