Claude review: drm/v3d: Use raw seqcount helpers instead of fighting with lockdep

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

Patch Review

The switch to `raw_write_seqcount_begin/end` and `raw_read_seqcount_begin` is justified by the commit message — jobs are serialized per queue, so there's no real race, and the lockdep annotations were producing false positives.

Minor nit: The read side changes `read_seqcount_begin` to `raw_read_seqcount_begin` but leaves `read_seqcount_retry` unchanged:

```c
	do {
		seq = raw_read_seqcount_begin(&stats->lock);
		...
	} while (read_seqcount_retry(&stats->lock, seq));
```

This is functionally correct since the lockdep annotation is on the `begin` side, not the `retry` side (which only has KCSAN annotations). But for consistency, you might consider whether pairing with `raw_read_seqcount_retry` (if available) would be cleaner.

The removal of the `IS_ENABLED(CONFIG_LOCKDEP)` conditional IRQ save/restore significantly simplifies the code. The comment in `v3d_drv.h` explaining why raw helpers are used is helpful.

---
Generated by Claude Code Patch Reviewer