Make close-issue command AI provider agnostic and CLI-callable

## Problem
The current `close-issue` command is tightly coupled to Claude Code's interactive shell environment and cannot be:
1. Called from the command line outside of Claude Code
2. Used with other AI providers (Amazon Q, Cursor, etc.)

## Solution
Create a provider-agnostic CLI tool that can be invoked both inside and outside interactive AI sessions.

## Implementation Plan

### 1. Create Standalone CLI Script (`bin/close-issue`)
- Shell script callable from command line: `close-issue 123`
- Support multiple AI providers via environment variables or flags
- Work both inside and outside Claude Code interactive sessions
- Use `claude -p` for non-interactive execution when using Claude

### 2. Extract Core Logic to Provider-Agnostic Module
- Move prompt generation from `.claude/command-templates/close-issue.md`
- Create `utils/close-issue-prompt.py` that:
- Aggregates knowledge base (like `.github/scripts/aggregate-knowledge.sh`)
- Generates the full prompt with injected procedures
- Returns formatted prompt for any AI provider

### 3. Update Template System for Multi-Provider Support
- Modify `utils/generate-commands.sh` to generate provider-specific wrappers
- Each provider gets a thin wrapper that calls the core CLI script
- Remove Claude-specific logging from core logic

### 4. Create Provider Abstraction Layer
- Support providers: Claude, Amazon Q, Cursor, generic LLM CLI
- Provider detection via environment variables or config
- Fallback chain: Claude → Amazon Q → Cursor → generic

### 5. Implementation Details
- **bin/close-issue**: Main CLI entry point
- Parse arguments (issue number, optional provider)
- Set up worktree (git operations)
- Generate prompt via Python module
- Execute via appropriate AI provider
- **utils/close-issue-prompt.py**: Prompt generation
- Load and inject knowledge base
- Process template placeholders
- Return complete prompt text
- **Provider adapters**: Thin wrappers for each AI tool
- Claude: `claude -p "$PROMPT"`
- Others: Similar non-interactive invocations

### 6. Maintain Backward Compatibility
- Keep `/close-issue` slash command working
- Update it to call the new CLI script internally
- Preserve existing logging for interactive sessions

## Benefits
- ✅ AI provider agnosticism via abstraction layer
- ✅ CLI-callable outside interactive sessions
- ✅ Reusable core logic across providers
- ✅ Maintains existing slash command functionality
- ✅ Follows systems-stewardship principle (single source of truth)

## Principles
- **systems-stewardship**: Single source of truth for issue closing logic
- **ai-provider-agnosticism**: Work across different AI tools
- **backward-compatibility**: Preserve existing workflows

Start by reading the existing .claude/command-templates/close-issue.md, utils/generate-commands.sh, and .github/scripts/aggregate-knowledge.sh to trace the current workflow. Then inspect how the /close-issue command sets up worktrees, generates prompts, and invokes Claude. Done means bin/close-issue works from and outside interactive sessions, supports the listed provider paths, and preserves the existing slash command behavior.

Written by the indexing model from the issue text.