Refined llms.txt. Added GkeCodeExecutor

[alexlam0206]

• Agent Configuration: Added details on LlmAgent's planner and code_executor attributes, and introduced the declarative Agent Config feature.
• Tool Confirmation: Included a new section explaining the tool confirmation flow for user verification and structured data input.
• State Management: Clarified the mechanisms for accessing state in instructions ({key} templating) and the recommended patterns for updating state via output_key, EventActions.state_delta, and context objects.
• Callbacks: Added a detailed section on Callbacks, explaining their lifecycle points, context objects, and use in controlling execution flow.
• GKE Code Executor: Added GkeCodeExecutor to the list of built-in tools, describing its use for secure, sandboxed code execution.