Add agent-audit skill for Cortex Agent best-practices auditing

[Resheras]

Summary

• Adds a new community skill agent-audit (invoked as $agent-audit) for auditing Snowflake Cortex Agent configurations
• Covers a 22-point checklist across six categories: Correctness & Hygiene, Instruction Architecture, Audience & Domain Context, Behavioral Controls, Response Quality, and Infrastructure & Robustness
• Includes optional Semantic View quality checks for agents using Cortex Analyst

Test plan

• Invoke $agent-audit Review my Cortex Agent for best practices in a Cortex Code session
• Confirm the skill extracts the full agent_spec and runs all 22 checks
• Confirm findings are grouped by severity: Critical / High / Nice-to-have