feat(worker): Ability to partition tasks per GPU

[JannikSt]

No description provided.

[Copilot]

Pull Request Overview

This PR adds GPU partitioning support for tasks by extending container naming and orchestration. Key updates include:

• Adding an optional gpu_index field and GPU suffix parsing in TaskContainer.
• Updating DockerService to handle multiple container creation and log aggregation per GPU.
• Adjusting task and container restart logic to support partitioned tasks.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
crates/worker/src/docker/task_container.rs	Added gpu_index field and updated parsing to extract GPU suffix from container names.
crates/worker/src/docker/service.rs	Implemented new logic to generate container names per GPU, start containers with GPU-specific environment variables, and aggregate container logs.
crates/shared/src/models/task.rs	Added partition_by_gpu flag to Task and TaskRequest models.

Comments suppressed due to low confidence (2)

crates/worker/src/docker/task_container.rs:27

• [nitpick] Consider adding a brief comment explaining the expected container name format and the rationale for checking the GPU suffix to improve code clarity.

            let (config_hash, gpu_index) =

crates/worker/src/docker/service.rs:300

• [nitpick] Refactor the nested check for gpu_index inside the GPU partitioning block to simplify the conditional logic and improve readability.

                                            let container_gpu = if payload.partition_by_gpu && gpu_index.is_some() {