fix: prevent KeyError when clearing and re-rendering DataTable

[xbrxr03]

Summary

Fixes #4900

DataTable.clear() was not clearing _updated_cells and _new_rows, leaving stale RowKey references that could cause a KeyError when the table was re-rendered after clearing and adding new rows.

Root Cause

When DataTable.clear() is called, it clears self.rows, self._data, and self._row_locations, but it does not clear self._updated_cells or self._new_rows. These sets retain stale RowKey/CellKey objects that reference rows which no longer exist in the table.

Since RowKey(value=None) uses id(self) as its hash, these stale key objects will never match any key in the rebuilt self.rows dict. When the on_idle handler processes these stale references, it can trigger a KeyError in code paths that access self.rows[row_key] directly (without .get()).

Notably, remove_row() already cleans up _updated_cells for the removed row — clear() should do the same for all rows.

Fix

Add two lines to DataTable.clear() to clear the stale reference sets:

self._updated_cells.clear()
self._new_rows.clear()

Test

Added test_clear_clears_updated_cells_and_new_rows which verifies that both _updated_cells and _new_rows are empty after calling clear().

Checklist

• Fix is minimal — only adds 2 lines
• Tests pass (88/88)
• Regression test added