Update skill documentation

[atharvacoolkni]

This pull request updates the documentation for the skills system to reflect a new, simplified approach to skill activation and resource loading. The most important changes are the removal of support and documentation for multiple active skills and the clear_skill tool, the introduction of on-demand resource loading via the set_skill tool, and a rework of how the agent and LLM interact with skills and resources. The documentation is now clearer about how skills and resources are used, and the system is described as stateless, with all skill content returned directly as tool results.

Skills API and Behavior Changes:

• Removed support for multiple active skills and the auto_deactivate option; skills are now activated one at a time, and the clear_skill tool is no longer documented. (docs/reference/library/skills/api-reference.md, docs/Tutorial/skills.md) [1] [2] [3]
• The agent now registers only the set_skill tool (not clear_skill), and all references to multiple active skills and explicit deactivation are removed. (docs/Tutorial/skills.md, docs/reference/library/skills/index.md) [1] [2]

Resource Loading and Usage:

• Added support for on-demand resource loading: the LLM can call set_skill(skill_name, resource_name) to load a specific resource file, and the tool returns the resource content directly. (docs/Tutorial/skills.md, docs/reference/library/skills/api-reference.md, docs/reference/library/skills/skill-format.md) [1] [2] [3]
• Updated documentation to describe how resources are listed in the trigger table and how the LLM should request them. (docs/Tutorial/skills.md, docs/reference/library/skills/skill-format.md) [1] [2]

System Prompt and State Handling:

• The system prompt now includes a trigger table listing all skills and their resources, and skill activation returns the full skill content as a tool result, which the LLM uses directly—no state is stored in execution_meta. (docs/reference/library/skills/api-reference.md, docs/reference/library/skills/index.md) [1] [2] [3]

Documentation Improvements:

• Clarified the skill body and resource loading process, emphasizing that skill content is returned as tool results and not injected or persisted in state. (docs/reference/library/skills/skill-format.md) [1] [2]
• Updated all examples and workflow diagrams to match the new stateless, tool-result-driven approach. (docs/Tutorial/skills.md, docs/reference/library/skills/index.md) [1] [2]

These changes make the skills system simpler, more robust to context trimming, and easier for both developers and LLMs to use.