sql-cli 1.68.0

SQL query tool for CSV/JSON with both interactive TUI and non-interactive CLI modes - perfect for exploration and automation
Documentation
# Input Migration Status

## Completed ✅

### Data Fields Removed
- `csv_client`, `csv_mode`, `csv_table_name` - removed from TUI
- `cache_mode`, `cached_data` - removed from TUI  
- `filtered_data` - removed from TUI
- All data now exclusively accessed through Buffer

### Wrapper Methods Updated
- All data access methods now require buffer (no fallbacks)
- `is_csv_mode()`, `get_csv_table_name()`, etc. - buffer-only
- `get_cached_data()`, `has_cached_data()` - buffer-only
- `get_filtered_data()`, `set_filtered_data()` - buffer-only

### Architecture Improvements
- TUI no longer holds duplicate state for data
- Single source of truth for all data (Buffer)
- Sorting refactored to work with immutable data

## In Progress 🚧

### Input Field Migration
- `input: Input` field removed from struct definition
- `textarea: TextArea` field removed from struct definition  
- `edit_mode: EditMode` field removed from struct definition
- Wrapper methods updated to use buffer's InputManager

### Current Issues
1. Many direct references to `self.input` and `self.textarea` throughout the code
2. Special modes (Search, Filter, etc.) need different handling
3. Complex interactions with vim mode, completion, etc.

## Remaining Work 📝

### Phase 11 Completion
1. **Fix all compilation errors** from removed input fields
   - ~20+ references to `self.textarea.*`
   - Several references to `self.input.*`
   - Buffer switching code that tries to set these fields

2. **Route all input through BufferAPI**
   - `handle_command_input()` needs to use buffer's InputManager
   - `handle_vim_input()` needs buffer integration
   - Completion system needs buffer integration

3. **Handle special modes**
   - Search/Filter modes need their own input handling
   - Could use a temporary input field just for these modes
   - Or create mini-buffers for each mode

### Recommended Approach

Given the complexity, consider:

1. **Option A: Incremental Migration**
   - Keep `input` field temporarily for special modes only
   - Route Command/Results modes through buffer
   - Gradually migrate special modes

2. **Option B: Complete Refactor**
   - Create InputState enum with variants for each mode
   - Each variant holds its appropriate input mechanism
   - Full separation of concerns

3. **Option C: Mini-Buffer System**
   - Create lightweight buffers for Search/Filter modes
   - Unify all input handling through buffer system
   - Most consistent but requires more work

## Benefits Achieved So Far

- ✅ Removed all data duplication between TUI and Buffer
- ✅ Single source of truth for all data state
- ✅ TUI is much lighter weight
- ✅ Better separation of concerns
- ✅ Foundation for DataTable abstraction

## Next Steps

1. Decide on approach for special modes input handling
2. Fix compilation errors systematically
3. Test thoroughly with all input modes
4. Complete Phase 11 of migration plan