Set the inference server port based on SAGEMAKER_BIND_TO_PORT environment variable

### 🚀 The feature, motivation and pitch

Set the inference server port based on `SAGEMAKER_BIND_TO_PORT` environment variable, to support multi-container endpoints and inference pipelines.

If the container needs to listen on a second port, choose a port in the range specified by the SAGEMAKER_SAFE_PORT_RANGE environment variable.

Reference:
https://docs.aws.amazon.com/sagemaker/latest/dg/multi-container-troubleshooting.html#multi-container-missing-accept

### Use case description

_No response_

### Alternatives

_No response_

### Additional context

_No response_

Start with the AWS SageMaker multi-container troubleshooting reference linked in the issue and trace how the inference server currently chooses its port. Verify behavior for SAGEMAKER_BIND_TO_PORT and the SAGEMAKER_SAFE_PORT_RANGE second-port case; done means the container listens on the requested port and selects a safe port when needed.

Written by the indexing model from the issue text.