fix: add flexible Json type annotation for Pydantic models (#1597)

[Sanjeev888]

What kind of change does this PR introduce?

Bug fix

What is the current behavior?

PostgREST returns JSON/JSONB columns as already-deserialized Python objects (dict, list, int, etc.). When validating these objects against Pydantic models typed with pydantic.Json, Pydantic raises a ValidationError because pydantic.Json requires raw string input.

Fixes #1597

What is the new behavior?

Added a flexible, subscriptable Json type annotation in postgrest.types (re-exported in supabase and supabase.types) using pydantic.BeforeValidator. It accepts both already-deserialized Python data objects and raw JSON-encoded strings.

Usage

from supabase import Json

class MyTable(BaseModel):
    id: int
    data: Json                   # accepts dict, list, scalar, or json string
    typed_data: Json[MySubModel] # parses json string if needed, validates as MySubModel