Update Dockerfile

### Changes - Update to non-[deprecated ](https://gitlab.com/nvidia/container-images/cuda/-/blob/master/doc/support-policy.md) base image version - Preinstall `deepspeed` in Docker image so this isn't done each time the app is run - Optimize Docker layer caching by installing `xtts-api-server` after `torch` & `torchaudio` (which likely change less frequently) - Listen on all network interfaces, ensuring the server is reachable from the host machine - Set the tunnel URL argument in the Docker `CMD` to localhost so the `preview_urls` returned by `/speakers` are accessible from the host machine <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit - **New Features** - Upgraded to a newer CUDA version for enhanced performance. - Integrated DeepSpeed for optimized model training. - **Refactor** - Updated the server command to improve API responsiveness and consistency. <!-- end of auto-generated comment: release notes by coderabbit.ai -->