Claude review: drm/damage-helper: Test src coord in drm_atomic_helper_check_plane_damage()

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

Moves the source-coordinate comparison from the runtime damage iterator into the check phase:

```c
+	if (!drm_rect_equals(&new_plane_state->src, &old_plane_state->src))
+		new_plane_state->ignore_damage_clips = true;
```

This is the right place for it -- alongside the modeset check. The function signature now takes both old and new state:
```c
void drm_atomic_helper_check_plane_damage(struct drm_atomic_commit *state,
+					  const struct drm_plane_state *old_plane_state,
+					  struct drm_plane_state *new_plane_state)
```

The corresponding removal from `drm_atomic_helper_damage_iter_init()`:
```c
-	if (!iter->clips || state->ignore_damage_clips ||
-	    !drm_rect_equals(&state->src, &old_state->src)) {
+	if (!iter->clips || state->ignore_damage_clips) {
```

Note: the `old_state` parameter is left in `damage_iter_init` for now but documented as "Unused, pass NULL". This keeps the series bisectable.

The deletion of 6 kunit tests that specifically tested src-moved scenarios is appropriate since that logic is now consolidated into the `ignore_damage_clips` flag, and the new test from patch 1 covers the flag behavior. However, the removed `drm_test_damage_iter_damage_not_visible` test checked the `visible = false` case combined with src movement -- the `visible = false` case by itself is still covered by `drm_test_damage_iter_no_damage_not_visible`, so this is fine.

**No issues.**

---
Generated by Claude Code Patch Reviewer