Fix mypy arg-type errors in generated discriminated union encoders

The codegen for discriminated union TypedDict encoders generates a ternary
chain like:

    encode_Foo(x) if x['kind'] == 'foo' else encode_Bar(x)

mypy cannot narrow TypedDict union types through key-based ternary conditions,
so every branch (not just the last fallback) receives the full union type.
The multi-variant sub-discriminator path (len(oneof_ts) > 1) already had
`# type: ignore[arg-type]` on line 306, but the single-variant path
(the common case) was missing it.

This caused the pid2 codegen CI in ai-infra to fail when the scribe schema
added new shape-type discriminated union variants.

Author: darshkpatel

src/replit_river/codegen/client.py

@@ -335,6 +335,7 @@ def flatten_union(tpe: RiverType) -> list[RiverType]:
                         # encoder_names.add(encoder_name)
                         typeddict_encoder.append(
                             f"{render_literal_type(encoder_name)}(x)"
+                            "  # type: ignore[arg-type]"
                         )
                     typeddict_encoder.append(
                         f"""

tests/v1/codegen/snapshot/snapshots/test_unknown_enum/enumService/needsEnumObject.py

@@ -71,9 +71,9 @@ def encode_NeedsenumobjectInput(
     x: "NeedsenumobjectInput",
 ) -> Any:
     return (
-        encode_NeedsenumobjectInputOneOf_in_first(x)
+        encode_NeedsenumobjectInputOneOf_in_first(x)  # type: ignore[arg-type]
         if x["kind"] == "in_first"
-        else encode_NeedsenumobjectInputOneOf_in_second(x)
+        else encode_NeedsenumobjectInputOneOf_in_second(x)  # type: ignore[arg-type]
     )