[Example Request]Hosting multiple booster models using SageMaker Multi-model endpoints on NVIDIA Triton

**Describe the use case example you want to see**
Customers are looking to hosting multiple booster models using NVIDIA Triton and leverage SageMaker's multi-model endpoints to dynamically load and unload the models to save costs.

**How would this example be used? Please describe.**
This example demonstrates how customers can bring their pre-trained XGBoost models and host them on CPU and GPU on Triton using SageMaker multi-model endpoints. Also, this examples demos how MME dynamically loads/unloads the models.

**Describe which SageMaker services are involved**
SageMaker Hosting

**Describe what other services (other than SageMaker) are involved***
SageMaker, S3, ECR, IAM

**Describe which dataset could be used. Provide its location in s3://sagemaker-sample-files or another source.**
N/A

Review existing SageMaker Hosting, NVIDIA Triton, and multi-model endpoint examples before choosing the notebook entry point. Done means an example shows pre-trained XGBoost models hosted on CPU and GPU with SageMaker multi-model endpoints, including dynamic model loading and unloading and the stated S3, ECR, and IAM services.

Written by the indexing model from the issue text.