Claude review: drm/i915/reset: Reorganize display reset code

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

Patch Review

This patch does two things: (1) extracts an `intel_display_reset_supported()` helper returning `HAS_DISPLAY(display)`, and (2) changes `intel_display_reset_prepare()` from returning `bool` to `void`, moving the `HAS_DISPLAY` guard into the caller (`intel_gt_reset_global`).

Looking at the caller in `intel_reset.c`, after this patch the `reset_display` flag is no longer conditionally updated by `intel_display_reset_prepare()`'s return value — it stays `true` once set, and `intel_display_reset_finish()` is always called when `reset_display` is true:

```c
if (reset_display)
    intel_display_reset_prepare(display,
                                display_reset_modeset_stuck,
                                gt);

intel_gt_reset(gt, engine_mask, reason);

if (reset_display)
    intel_display_reset_finish(display, !need_display_reset);
```

In the old code, if `intel_display_reset_prepare()` returned `false` (only happened when `!HAS_DISPLAY`), `reset_display` would be cleared and `intel_display_reset_finish()` wouldn't be called. Now with `HAS_DISPLAY` checked earlier via `intel_display_reset_supported()`, the `reset_display` flag correctly guards both prepare and finish.

However, `intel_display_reset_finish()` still has its own `if (!HAS_DISPLAY(display)) return;` guard in the current tree version, which this patch removes. That's fine because the caller already gates on `intel_display_reset_supported()`.

But there's a subtle behavioral change: previously when `intel_display_reset_prepare()` failed to duplicate state (returned `true` but set `modeset_state = NULL`), `intel_display_reset_finish()` would still be called and would drop locks. Now with `void` return type, the same thing happens — this is preserved correctly since the `if (!state) goto unlock;` path in `_finish` handles this.

**Looks correct.**

One minor nit: `intel_display_reset_test()` now also checks `HAS_DISPLAY(display)` redundantly — the caller already gates on `intel_display_reset_supported()`. Not a bug, just belt-and-suspenders.

---
Generated by Claude Code Patch Reviewer