Claude review: drm: renesas: rzg2l_mipi_dsi: Move rzg2l_mipi_dsi_set_display_timing()

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

Patch Review

This moves `rzg2l_mipi_dsi_set_display_timing()` from `atomic_pre_enable` to `atomic_enable`, placing it after `rzg2l_mipi_dsi_start_hs_clock()`. The hardware manual requires display timings to be set after HS clock is started.

**Issue: Silent dropping of error handling for rzg2l_mipi_dsi_startup()**

In the original code:
```c
	ret = rzg2l_mipi_dsi_startup(dsi, mode);
	if (ret < 0)
		return;
```

The patch changes this to:
```c
	rzg2l_mipi_dsi_startup(dsi, mode);
```

`rzg2l_mipi_dsi_startup()` returns an int and can fail (e.g., `pm_runtime_resume_and_get()`, `dphy_init()`, `dphy_conf_clks()` can all return errors). Silently ignoring the return value means that if startup fails, the driver will proceed to `atomic_enable` and attempt to start the HS clock on uninitialized hardware. The commit message mentions "Drop the unused ret variable" but `ret` was being *used* for error checking — it just couldn't propagate the error since the bridge callback returns void. Still, the early return prevented further damage. This should be preserved:

```c
	if (rzg2l_mipi_dsi_startup(dsi, mode) < 0)
		return;
```

**Duplicated connector/crtc/mode lookup:** The mode lookup code is now duplicated in both `atomic_pre_enable` and `atomic_enable`. Consider caching the mode pointer in `struct rzg2l_mipi_dsi` during `atomic_pre_enable` to avoid the duplication, though this is a minor style point.

Otherwise the fix itself is correct per the hardware manual requirements.

---
Generated by Claude Code Patch Reviewer