Refactor: Split handleDragEnd into column/card sub-handlers

[ryota-murakami]

Background

Code quality analysis (2026-02-20) identified handleDragEnd in KanbanBoard as an 80+ line function with deeply nested conditionals that harms readability.

Current State

handleDragEnd in src/components/Board/KanbanBoard.tsx handles both column and card drag operations in a single function:

const handleDragEnd = async (event: DragEndEvent) => {
  // Reset state
  if (!over) return

  // Column drag operations (~30 lines)
  if (active.data.current?.type === COLUMN_DRAG_TYPE) {
    if (new row) → handleNewRowDrop
    if (insert zone) → handleColumnInsertDrop  
    if (column swap) → handleColumnSwap
    return
  }

  // Card drag operations (~50 lines, inline logic)
  // - Resolve target column
  // - Save history
  // - Same-column reorder vs cross-column move
  // - DB sync with error handling
}

The column branch already delegates to sub-handlers. The card branch has ~50 lines of inline logic.

Proposed Change

Extract handleCardDragEnd as a separate useCallback:

const handleCardDragEnd = useCallback(async (active, over) => {
  // Card-specific logic here
}, [cards, cardsByStatus, dispatch, EMPTY_CARDS])

const handleDragEnd = async (event: DragEndEvent) => {
  setActiveId(null)
  setActiveDragType(null)
  if (!over) return

  if (active.data.current?.type === COLUMN_DRAG_TYPE) {
    // Column logic (already delegated)
    return
  }
  
  await handleCardDragEnd(active, over)
}

Acceptance Criteria

• handleCardDragEnd extracted as a separate useCallback
• handleDragEnd reduced to ~20 lines (dispatch + delegation)
• No behavior changes — pure readability improvement
• All existing tests pass

Notes

• This can be done standalone, but fits naturally into Issue Refactor: Extract useKanbanDnD hook from KanbanBoard (God Component decomposition) #141 (useKanbanDnD extraction)
• If Refactor: Extract useKanbanDnD hook from KanbanBoard (God Component decomposition) #141 is implemented first, this becomes part of that hook's internal structure