Fix incorrect construction of LabwareOffsetVectors.

The old code was type-unsafe (`cast()`) but we got away with it because both types had compatible fields. Earlier commits in this PR tried to fix that by using `model_validate()` instead of `cast()`. But `model_validate()` can't directly convert one Pydantic model into another like that, so it raised a runtime error. That wasn't caught by the type checker because `model_validate()` is, sensibly, typed to allow any untrusted input.

This converts between the two types verbosely and type-safely, by spelling out each field.

api/src/opentrons/protocol_engine/state/labware.py

@@ -998,11 +998,15 @@ def get_child_gripper_offsets(
             return None
         else:
             return LabwareMovementOffsetData(
-                pickUpOffset=LabwareOffsetVector.model_validate(
-                    parsed_offsets[offset_key].pickUpOffset
+                pickUpOffset=LabwareOffsetVector.model_construct(
+                    x=parsed_offsets[offset_key].pickUpOffset.x,
+                    y=parsed_offsets[offset_key].pickUpOffset.y,
+                    z=parsed_offsets[offset_key].pickUpOffset.z,
                 ),
-                dropOffset=LabwareOffsetVector.model_validate(
-                    parsed_offsets[offset_key].dropOffset
+                dropOffset=LabwareOffsetVector.model_construct(
+                    x=parsed_offsets[offset_key].dropOffset.x,
+                    y=parsed_offsets[offset_key].dropOffset.y,
+                    z=parsed_offsets[offset_key].dropOffset.z,
                 ),
             )