feat: Add UserFeedback tool for dynamic user interaction (#532)

[sudabg]

Description

Implements the UserFeedback tool as described in #532, enabling agents to request user input at decision points while operating in autonomous mode.

What's Added

superagi/tools/user_feedback/:

• tools.py — UserFeedbackTool class with schema and execution logic
• prompts/user_feedback.txt — Prompt template for formatting feedback requests
• __init__.py — Package init

Tool Behavior

1. Agent identifies a need for user input (ambiguity, confirmation, missing context)
2. Agent calls UserFeedbackTool with a question and optional context
3. Tool formats a clear feedback request message
4. Message is logged and returned to the agent execution feed
5. SuperAGI framework presents it to the user and captures response

Use Cases

• Decision points: "Should I proceed with deploying to production?"
• Ambiguous requirements: "What time zone should I use for the deadline?"
• Safety gates: "Confirm you want to delete these 50 files?"
• Missing context: "What is the API endpoint URL?"

Integration

Works alongside existing modes:

• God Mode: Agent calls tool autonomously at key milestones
• Restricted Mode: Complements permission checks with proactive questions
• Human-in-the-Loop: Natural pause points for user guidance

Example

# Agent execution
result = UserFeedbackTool()._execute(
    question="Which database schema should I use?",
    context="Found 3 possible schemas: users_v1, users_v2, users_v3"
)
# Returns formatted request awaiting user response

Closes #532