Update version pin to onnx 1.16

[stellaraccident]

onnx 1.16 includes support for Python 3.12. We should upgrade. Based on a user report, there may need to be some cleanup.

[zjgarvey]

Just to verify, you would like me to

1. pin onnx==1.16.0 in test-requirements.txt
2. check for e2e test regressions with config onnx and fix them
3. setup building nightly python wheels for python version 3.12
4. anything else?

[stellaraccident]

That's it.

[zjgarvey]

Sounds good. Thanks.

[stellaraccident]

FYI - making this change and then ninja check-torch-mlir yields issues.

Here is a subset:

======================================================================                                                                                                                              ERROR: test_node_test_cast_INT4_to_FLOAT16_model (__main__.ImportSmokeTest.test_node_test_cast_INT4_to_FLOAT16_model)                                                                               ----------------------------------------------------------------------                                                                                                                              Traceback (most recent call last):                                                                                                                                                                    File "/home/stella/src/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/extras/onnx_importer.py", line 484, in tensor_element_type                                             t = ELEM_TYPE_TO_IR_TYPE_CB[elem_type]()                                                                                                                                                                ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^                                                                                                                                                          KeyError: 22                                                                                                                                                                                                                                                                                                                                                                                            During handling of the above exception, another exception occurred:                                                                                                                                                                                                                                                                                                                                     Traceback (most recent call last):                                                                                                                                                                    File "/home/stella/src/torch-mlir/test/python/onnx_importer/import_smoke_test.py", line 128, in test_method                                                                                           raise e                                                                                                                                                                                           File "/home/stella/src/torch-mlir/test/python/onnx_importer/import_smoke_test.py", line 121, in test_method                                                                                           self.run_import_test(norm_name, rel_path)                                                                                                                                                         File "/home/stella/src/torch-mlir/test/python/onnx_importer/import_smoke_test.py", line 99, in run_import_test                                                                                        imp = onnx_importer.NodeImporter.define_function(model_info.main_graph, m)                                                                                                                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^                                                                                                                        File "/home/stella/src/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/extras/onnx_importer.py", line 205, in define_function                                                 cc.type_proto_to_type(inp.type) for inp in graph_info.input_map.values()                                                                                                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^                                                                                                                                                                   File "/home/stella/src/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/extras/onnx_importer.py", line 590, in type_proto_to_type                                              element_type = self.tensor_element_type(tt.elem_type)                                                                                                                                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^                                                                                                                                             File "/home/stella/src/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/extras/onnx_importer.py", line 486, in tensor_element_type                                             raise OnnxImportError(f"Unknown ONNX tensor element type: {elem_type}")                                                                                                                         torch_mlir.extras.onnx_importer.OnnxImportError: Unknown ONNX tensor element type: 22

Some others that look related. Seems like some API/type adaptation needed.