Migrating optimizer to new IR (part 1) (#1725)

Migrated the core logic. 

TO DO: decide how to handle functions. Optimizer currently incorporates
function-specialization. Need to choose between function-specialization
and function-inlining.

onnxscript/ir/serde.py

@@ -50,6 +50,7 @@
     "serialize_tensor_into",
     "serialize_tensor",
     "serialize_type_into",
+    "serialize_type",
     "serialize_value_into",
     "serialize_value",
     "SerdeError",
@@ -1511,6 +1512,12 @@ def serialize_type_into(type_proto: onnx.TypeProto, from_: _protocols.TypeProtoc
         raise TypeError(f"Unsupported type: {from_}")
 
 
+def serialize_type(type_protocol: _protocols.TypeProtocol) -> onnx.TypeProto:
+    type_proto = onnx.TypeProto()
+    serialize_type_into(type_proto, from_=type_protocol)
+    return type_proto
+
+
 @_capture_errors(lambda type_proto, from_: repr(from_))
 def serialize_shape_into(type_proto: onnx.TypeProto, from_: _protocols.ShapeProtocol) -> None:
     value_field = type_proto.WhichOneof("value")