xlsx: error on cell references beyond the grid limits

[krickert]

Fixes #694.

Both accumulators in get_row_and_optional_column were unchecked u32, so a reference like ZZZZZZZ1 or A99999999999 panicked in debug builds and wrapped in release builds, yielding cell positions that appear nowhere in the file.

As requested in review, the fix rejects invalid references instead of guarding the arithmetic: row and column accumulation is checked, and a reference past the grid (row > 1,048,576 or column > 16,384) returns the existing XlsxError::RowNumberOverflow / ColumnNumberOverflow. Dimension parsing propagates the same errors, replacing the warn-and-continue path in get_dimension. The column step moved into push_column_letter so the uppercase and lowercase arms stay identical.

test_reference_beyond_grid_is_an_error covers one past the grid on each axis, references that overflow u32 entirely, the same through get_dimension, and the bottom-right cell XFD1048576, which still parses.

Touches the same get_dimension arm as #695; whichever lands second gets rebased.

[jmcnamara]

This PR is fixing the wrong thing. Rather than handling U32 overflow it should be handling invalid row/column dimensions and raising an error.

[krickert]

Done.
Reworked the fix in 15fad0f: get_row_and_optional_column now validates against the grid and returns XlsxError::RowNumberOverflow (row > 1,048,576) or XlsxError::ColumnNumberOverflow (column > 16,384) instead of only guarding the u32 bounds.

The lenient warn! path in get_dimension is gone so dimension parsing propagates the error. The checked arithmetic is retained for the beyond-u32 case.

test_reference_beyond_grid_is_an_error covers one-past-grid on both axes, u32 overflow, and the in-bounds corner XFD1048576.

[krickert]

Reworked to reject invalid references as requested, let me know what you think.