Cannot register custom DML operator

@RandySheriffH is already working on this.

Since Jun 26, 2023.

### Describe the issue

Hello,

I'm attempting to register a custom operator for the DmlExectionProvider, and getting a crash from `GraphPartitioner.cpp`.

The message is: `Assertion failed: createInfo != nullptr, file \core\providers\dml\DmlExecutionProvider\src\GraphPartitioner.cpp, line 166`.

```c++
struct MyCustopOpKernel { ... };
struct MyCustomOp : Ort::CustomOpBase { ... }

Ort::CustomOpDomain custom_op_domain("my_domain");
MyCustomOp custom_op{};
custom_op_domain.Add(&custom_op);

Ort::SessionOptions sessionOptions;
sessionOptions.Add(custom_op_domain);
// ... other DML-specific session options configurations..
ortDmlApi->SessionOptionsAppendExecutionProvider_DML1(sessionOptions, dmlDevice, dmlQueue);
// crash occurs on this line:
ortSession = std::make_unique(ortEnv, modelData.data(), modelData.size(), sessionOptions);
```

I know that onnxruntime is able to recognize my op declaration, because if the domain name or the op name don't match what's in the model, I get a helpful message letting me know that the op/function was not found. I when correctly configured, it gets past that point, only to crash deeper in the Ort::Session initialization. Based on the where the crash happens in the ort source code, it seems like I need to access the `RegisterDmlOperators` function from `DmlExecutionProvider.h` file. But its not clear that this is part of the public API? It's not exported from `dml_provider_factory.h`, which I had been using as my entry point to the onnxruntime DirectML functionality. So I'm wondering if implementing custom DML operator is even supported? And if so, what am I missing here?

### To reproduce

I am happy to upload a more complete repro but I wanted to get a confirmation that what I'm trying to do is even supported first :)

### Urgency

:)

### Platform

Windows

### OS Version

10

### ONNX Runtime Installation

Built from Source

### ONNX Runtime Version or Commit ID

latest (1.15)

### ONNX Runtime API

C++

### Architecture

X64

### Execution Provider

DirectML

### Execution Provider Library Version

_No response_