Merge feat/buffer-system: Phase 1 buffer migration complete

This merge brings in significant improvements to the TUI architecture:

Buffer System Migration (Phase 1):
- Introduced BufferAPI trait for abstraction
- Created Buffer struct with independent state management
- Added BufferManager for multi-buffer support
- Migrated simple fields (edit_mode, case_insensitive, etc.)
- Added buffer debug dump to F5 output

Fixes and Improvements:
- Fixed case_insensitive icon not showing in status line
- Fixed column search text not rendering properly
- Fixed fuzzy filter and search modes to preserve SQL query
- Fixed fuzzy filter status line indicator persistence
- Added command history deduplication
- Auto-clean history file on startup
- Comprehensive test suite for complex queries

The buffer system now provides a solid foundation for:
- Multiple file/tab support
- Proper state isolation between buffers
- Easier testing and maintenance
- Future migration of complex state (Phase 2)

All search/filter modes now correctly preserve the SQL query
and provide proper visual feedback.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: TimelordUK

column-search.cast

@@ -0,0 +1,141 @@
+# Buffer Migration Audit
+
+This document tracks the migration of fields from EnhancedTuiApp to Buffer via BufferAPI.
+
+## Migration Categories
+
+### ✅ Already in Buffer (via BufferAPI)
+These fields already exist in Buffer and have BufferAPI methods:
+
+- [x] `mode: AppMode` → `get_mode()`, `set_mode()`
+- [x] `status_message: String` → `get_status_message()`, `set_status_message()`
+- [x] `results: Option<QueryResponse>` → `get_results()`, `set_results()`
+- [x] `table_state: TableState` → `get_selected_row()`, `set_selected_row()`
+- [x] `input: Input` → `get_input_value()`, `set_input_value()`, etc.
+- [x] `current_column: usize` → `get_current_column()`, `set_current_column()`
+- [x] `scroll_offset: (usize, usize)` → `get_scroll_offset()`, `set_scroll_offset()`
+- [x] `pinned_columns: Vec<usize>` → `get_pinned_columns()`, `add_pinned_column()`, etc.
+- [x] `compact_mode: bool` → `is_compact_mode()`, `set_compact_mode()`
+- [x] `show_row_numbers: bool` → `is_show_row_numbers()`, `set_show_row_numbers()`
+- [x] `filter_state: FilterState` → `get_filter_pattern()`, `is_filter_active()`, etc.
+- [x] `search_state: SearchState` → `get_search_pattern()`, `get_search_matches()`, etc.
+- [x] `sort_state: SortState` → `get_sort_column()`, `get_sort_order()`, etc.
+- [x] `csv_client: Option<CsvApiClient>` → `get_csv_client()`, `get_csv_client_mut()`
+- [x] `csv_mode: bool` → `is_csv_mode()`
+- [x] `csv_table_name: String` → `get_table_name()`
+- [x] `filtered_data: Option<Vec<Vec<String>>>` → `get_filtered_data()`, `set_filtered_data()`
+
+### ✅ Phase 1: Simple Fields (COMPLETED)
+These are straightforward to migrate and test:
+
+- [x] `edit_mode: EditMode` - How the editor is being used (single/multi-line)
+- [x] `last_results_row: Option<usize>` - Position preservation
+- [x] `last_scroll_offset: (usize, usize)` - Position preservation
+- [x] `case_insensitive: bool` - Search/filter behavior
+- [x] `last_query_source: Option<String>` - Track where query came from
+
+### 🔄 Phase 2: Buffer-Specific Complex State
+These need to move but require more care:
+
+- [ ] `textarea: TextArea<'static>` - Multi-line editor state
+- [ ] `fuzzy_filter_state: FuzzyFilterState` - Fuzzy search state
+- [ ] `column_search_state: ColumnSearchState` - Column search functionality
+- [ ] `column_widths: Vec<u16>` - Display formatting
+- [ ] `column_stats: Option<ColumnStatistics>` - Statistics for current column
+- [ ] `cached_data: Option<Vec<serde_json::Value>>` - Cached query results
+- [ ] `cache_mode: bool` - Whether caching is active
+- [ ] `undo_stack: Vec<(String, usize)>` - Undo history
+- [ ] `redo_stack: Vec<(String, usize)>` - Redo history
+- [ ] `kill_ring: String` - Vim-style yank buffer
+- [ ] `viewport_lock: bool` - Display behavior
+- [ ] `viewport_lock_row: Option<usize>` - Display behavior
+- [ ] `jump_to_row_input: String` - Jump command input
+- [ ] `last_visible_rows: usize` - Display tracking
+
+### 🚫 Should Stay in TUI (Global State)
+These are truly global and shouldn't be in buffers:
+
+- `api_client: ApiClient` - Global API connection
+- `sql_parser: SqlParser` - Global parser
+- `hybrid_parser: HybridParser` - Global parser
+- `config: Config` - User configuration
+- `command_history: CommandHistory` - Global command history
+- `sql_highlighter: SqlHighlighter` - Syntax highlighting
+- `query_cache: Option<QueryCache>` - Global cache
+- `show_help: bool` - UI state
+- `help_scroll: u16` - UI state
+- `debug_text: String` - Debug UI
+- `debug_scroll: u16` - Debug UI
+- `input_scroll_offset: u16` - UI display
+- `selection_mode: SelectionMode` - Global UI preference
+- `yank_mode: Option<char>` - Global clipboard state
+- `last_yanked: Option<(String, String)>` - Global clipboard
+- `completion_state: CompletionState` - UI helper
+- `history_state: HistoryState` - UI helper
+- `buffer_manager: Option<BufferManager>` - The manager itself
+- `current_buffer_name: Option<String>` - Current buffer tracking
+
+## Migration Strategy
+
+### Step 1: Create Compatibility Layer
+For each field we want to migrate, create a method that:
+1. Checks if buffer_manager exists
+2. If yes, use BufferAPI
+3. If no, use direct field access
+
+Example:
+```rust
+fn get_edit_mode(&self) -> EditMode {
+    if let Some(buffer) = self.get_current_buffer() {
+        buffer.get_edit_mode()
+    } else {
+        self.edit_mode
+    }
+}
+```
+
+### Step 2: Update All References
+Search for all direct field accesses and replace with compatibility methods:
+- `self.edit_mode` → `self.get_edit_mode()`
+- `self.edit_mode = x` → `self.set_edit_mode(x)`
+
+### Step 3: Test After Each Migration
+Run the TUI and verify:
+1. Basic navigation works
+2. Editing works
+3. Filtering/searching works
+4. No crashes or unexpected behavior
+
+### Step 4: Move Field to Buffer
+Once all references use the compatibility layer:
+1. Add field to Buffer struct
+2. Add to BufferAPI trait
+3. Implement in Buffer
+4. Remove from EnhancedTuiApp
+
+## Progress Tracking
+
+### Phase 1 Checklist (Simple Fields)
+- [ ] Add `edit_mode` to BufferAPI
+- [ ] Create `get_edit_mode()` / `set_edit_mode()` compatibility methods
+- [ ] Update all `self.edit_mode` references
+- [ ] Test TUI functionality
+- [ ] Move field to Buffer
+- [ ] Repeat for other Phase 1 fields
+
+### Testing Checklist
+After each field migration:
+- [ ] Can open a CSV file
+- [ ] Can type and execute queries
+- [ ] Can navigate results
+- [ ] Can filter results
+- [ ] Can search in results
+- [ ] Can yank/copy data
+- [ ] No crashes or panics
+
+## Notes
+
+- We can merge to master at any point - the compatibility layer ensures nothing breaks
+- Each field migration is independent - we can do them one at a time
+- If something goes wrong, we can easily revert a single field migration
+- The goal is gradual, safe migration, not speed