Refactor public interface of completion module.

What does this merge request do and why?

• Clean up the public interfaces of the completion module: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/70#note_1566379801
• Change BatchRequestCompletionsVertex to be single model only: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/71#note_1564291700
• Use aiohttp instead of thread: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/68#note_1568527107
• Add dataclass for model metadata for cleaner mapping: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/70#note_1566370176
• Move model input token limit to ModelMeta: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/66#note_1571018836
• Avoid importing anything from prompt-lib in prompt-engine because prompt-lib should strictly depend on prompt-engine.

How to set up and validate locally

Numbered steps to set up and validate the change are strongly suggested.

Merge request checklist

• I've ran the eval_codebase.py pipeline to validate that nothing is broken. https://console.cloud.google.com/bigquery?project=unreview-poc-390200e5&ws=!1m5!1m4!4m3!1sunreview-poc-390200e5!2sgl_gitlab_experiments!3stest_huggingface_model
• Tests added for new functionality. If not, please raise an issue to follow up.
• Documentation added/updated, if needed.