Claude review: drm/rockchip: cdn-dp: Support handle lane info without extcon

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

Patch Review

Makes the extcon device optional in CDN-DP, falling back to `phy_get_bus_width()` for lane count.

**Issue: Potentially incorrect filter condition in probe**
```c
-       if (IS_ERR(extcon) || IS_ERR(phy))
+       if (IS_ERR(phy) || PTR_ERR(extcon) != -ENODEV)
            continue;
```
The original skips the port if either extcon OR phy is an error. The new code skips if phy is an error OR if `PTR_ERR(extcon) != -ENODEV`. But if `extcon` is **valid** (not an error), then `PTR_ERR(extcon)` is undefined behavior — it would cast a valid pointer to a long and compare it against `-ENODEV`. This condition should be:
```c
if (IS_ERR(phy) || (IS_ERR(extcon) && PTR_ERR(extcon) != -ENODEV))
    continue;
```
The intent is: skip if phy fails, or if extcon failed with something other than `-ENODEV` (meaning extcon is genuinely absent vs. a real error like `-EPROBE_DEFER`). The current code is **buggy for the case where extcon is valid**.

Wait — re-reading: extcon is fetched above with `extcon_find_edev_by_node`. If it returns a valid pointer, `PTR_ERR()` of a valid pointer is not `-ENODEV`, so the `continue` would fire, skipping valid ports. This is a real bug.

**Issue: `property.intval = 0` default for polarity**
```c
property.intval = 0;
if (port->extcon) {
    ret = extcon_get_property(port->extcon, EXTCON_DISP_DP,
                  EXTCON_PROP_USB_TYPEC_POLARITY, &property);
```
When extcon is absent, polarity defaults to 0, which assumes normal polarity. This seems reasonable but should be documented or verified against the hardware behavior when the Type-C controller is managing polarity via the PHY directly.

---
Generated by Claude Code Patch Reviewer