A support for GPU

[zapadinsky]

A new switch --gpus is introduced that is passed to the docker
It also tries to deduce if gpu is needed to run the ci based on tags. like saas-linux-medium-amd64-gpu-standard

---

Summary by cubic

Add GPU support for CI jobs via a new --gpus flag and automatic tag-based detection. When enabled, containers run with GPUs and the required IPC/ulimit settings.

• New Features
  • New --gpus CLI option (indexes or all), forwarded to Docker.
  • Auto-enables if a job tag includes gpu as a dash-delimited token (e.g., saas-linux-medium-amd64-gpu-standard), defaulting to --gpus all.
  • When GPUs are enabled, adds Docker flags: --gpus ..., --ipc=host, --ulimit memlock=-1, --ulimit stack=67108864.
  • No change for non-GPU jobs unless the flag or tag is used.

^{Written for commit baaca82. Summary will update on new commits. Review in cubic}

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}

[zapadinsky]

I've fixed typos and string constants formatting. Force-pushed the changes.