ai: add docker model runner.

[abiosoft]

Generated by Copilot

This pull request introduces a major refactor to the AI model management commands, adding support for selecting between multiple model runners (Docker and Ramalama), streamlining runner setup and provisioning, and improving configuration handling. The changes also update command documentation and argument parsing, and unify model runner logic for easier extensibility.

Model runner selection and configuration:

• Added support for selecting the AI model runner (docker, ramalama) via the --runner flag, instance config (modelRunner), and default value. The runner is now used throughout model commands for provisioning, argument building, and serving. [1] [2] [3]
• Updated the start command to accept a --model-runner flag, and propagate this value into instance configuration. [1] [2]

Command logic and extensibility:

• Refactored the model command to use runner-specific methods for validation, provisioning, argument building, and serving, replacing hardcoded Ramalama logic with a unified runner interface.
• Added a new model serve command for serving models via a web UI and OpenAI-compatible API, using the selected runner.

Documentation and minor improvements:

• Improved command documentation and examples to reflect runner selection and usage, and updated help text for clarity.
• Updated version reporting to use the new runner interface for fetching Ramalama version. [1] [2]