Claude review: media: meson: vdec: Cancel esparser work during teardown

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

Patch Review

**Good:** Adding `cancel_work_sync()` in teardown paths is the right approach.

**Issue — drain moved before hardware teardown:**
```c
cancel_work_sync(&sess->esparser_queue_work);

mutex_lock(&core->lock);
...
sess->status = STATUS_STOPPED;
mutex_unlock(&core->lock);

if (q->type != V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE) {
    if (old_status >= STATUS_RUNNING && codec_ops->drain)
        codec_ops->drain(sess);
}

if (old_status == STATUS_RUNNING || ...)
    ...
    vdec_poweroff(sess);  /* also calls drain() internally! */
```

There are two problems here:
1. `codec_ops->drain()` is now called **before** `vdec_poweroff()`, but `vdec_poweroff()` also calls `codec_ops->drain()` internally (line 122-123 of the original). So drain is called **twice** — once explicitly here, and once inside `vdec_poweroff()`.
2. The commit message says "Move codec_ops->drain() evaluation earlier in stop_streaming() using the status snapshot, so draining occurs before buffer cleanup." But the original code already drained before buffer cleanup (drain was in the `else` branch before `v4l2_m2m_dst_buf_remove()`). The move creates the double-drain issue.

**Issue — `cancel_work_sync` in `vdec_close`:** This is fine as defense-in-depth, since `v4l2_m2m_ctx_release()` should trigger `stop_streaming` which already cancels it (after this patch).

---
Generated by Claude Code Patch Reviewer