OnnxRuntimeGenAIChatClientOptions support DeepSeek-r1-Distill-14b

I am not sure if the OnnxRuntimeGenAIChatClientOptions of the following deepseek model is correct. The value of message.Role.Value is fixed to "user". The user prompt tag of this deepseek model is < | User | >.
model:https://huggingface.co/onnxruntime/DeepSeek-R1-Distill-ONNX/tree/main/deepseek-r1-distill-qwen-14B/gpu/gpu-int4-rtn-block-32
```
public static OnnxRuntimeGenAIChatClientOptions GetPhi4() => new()
{
StopSequences = ["<|system|>", "<|user|>", "<|assistant|>", "<|end|>"],
PromptFormatter = static (messages, _) =>
{
var prompt = new StringBuilder();
foreach (var message in messages)
foreach (var content in message.Contents.OfType())
prompt.Append($"<|{message.Role.Value}|>\n{content.Text}<|end|>\n");
return prompt.Append("<|assistant|>\n").ToString();
}
};

public static OnnxRuntimeGenAIChatClientOptions GetDeepSeek() => new()
{
StopSequences = ["<｜end▁of▁sentence｜>"],
PromptFormatter = static (messages, _) =>
{
var prompt = new StringBuilder();
foreach (var message in messages)
foreach (var content in message.Contents.OfType())
prompt.Append($"<｜begin▁of▁sentence｜><｜{message.Role.Value}｜>\n{content.Text}<｜end▁of▁sentence｜>\n");
return prompt.Append("<｜Assistant｜>\n").ToString();
}
};
```

Start with the OnnxRuntimeGenAIChatClientOptions entry point and the GetDeepSeek formatter shown in the issue. Compare its role and prompt tokens with the linked DeepSeek-R1-Distill model format. Done means the options produce the model's expected user and assistant tags and stop sequence.

Written by the indexing model from the issue text.