Add issue #121's ICH repro as a regression test

[JohnCampionJr]

#121 reported the classic mid-line ICH smear (an overlapping forward copy replicating the cell under the cursor to the end of the row). It does not reproduce on current main — the fix landed in #80, which gave CopyCellsFrom memmove semantics (

XTerm.NET/src/XTerm.NET/Buffer/BufferLine.cs

Lines 341 to 350 in 3760bf0

	public void CopyCellsFrom(BufferLine src, int srcCol, int destCol, int length, bool applyInReverse)
	{
	// Memmove semantics, decided HERE rather than trusted to the caller: copying within one
	// line with the destination ahead of the source re-reads cells it has already written
	// when it walks forward, and the shifted region degenerates into the first cell repeated
	// -- which is exactly what a mid-line insert looked like on screen (type into the middle
	// of a bash command line and the tail becomes one letter, over and over). Every caller
	// that shifts right within a line is safe now whatever it passes.
	if (ReferenceEquals(src, this) && destCol > srcCol && destCol < srcCol + length)
	applyInReverse = true;

) — and the published 2.0.0-rc005 binary passes the issue's repro too.

This just pins the issue's exact shape down with a named test: the cursor arrives mid-line by backspaces rather than CUP, and the gap can be wider than one cell. The count > 1 case exercises the overlap test inside CopyCellsFrom with a shift distance the existing count-1 regression tests from #80 never reach.

Full suite green: 2,114 passed.

Closes #121.

🤖 Generated with Claude Code