Claude review: drm/radeon: Consistently define pci_device_ids using named initializers

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

Patch Review

**Positive observations:**

1. The 1:1 line count (700 removed, 700 added) confirms no entries were accidentally added or dropped.

2. The `PCI_DEVICE()` macro usage is correct -- it sets `.vendor`, `.device`, `.subvendor=PCI_ANY_ID`, `.subdevice=PCI_ANY_ID`, which matches the old `{vendor, device, PCI_ANY_ID, PCI_ANY_ID, ...}` pattern exactly.

3. All old entries had `class=0, class_mask=0` (the 5th and 6th positional fields), so the implicit zero-initialization from the designated-initializer form loses nothing.

4. The array terminator change from `{0, 0, 0}` to `{ }` is correct -- both produce an all-zeros `struct pci_device_id` sentinel.

5. Spot-checking several entries confirms vendor IDs, device IDs, and `driver_data` flag combinations are preserved exactly.

**Minor observations (not blocking):**

- The patch triggers checkpatch line-length warnings (as noted in the cover letter). The new lines are shorter than or equal to the old ones, so this is acceptable. The author offers alternative line-wrapping styles; the current single-line form is the most readable for a device ID table.

- The file only defines `radeon_PCI_IDS` and is only consumed by `radeon_drv.c:253`. The other drivers that include `drm_pciids.h` (gma500, mgag200, amdgpu) use different macros not present in this file, so there's no cross-driver impact.

**No issues found.** This is a clean, mechanical conversion that improves readability and enables the planned `driver_data` anonymous union work.

---
Generated by Claude Code Patch Reviewer