iter17 α-20 (test): MIN_CAP_POOL 24 -> 11 to match kdirect

Quick discriminator: if pool depth affects rkvdec's per-codec state
machine, reducing libva's pool to kdirect's ~11 might change Bug 4/5/6
hashes. Reverts to 24 if test shows no change or regression.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/cap_pool.h

@@ -67,7 +67,15 @@
 
 #include <linux/videodev2.h>	/* for VIDEO_MAX_PLANES */
 
-#define MIN_CAP_POOL 24
+/*
+ * iter17 α-20: reduced from 24 to 11 as a test — matches kdirect
+ * (ffmpeg-v4l2request) CAPTURE pool depth on the same fixture. iter15
+ * ioctl-sequence diff showed kdirect issues CREATE_BUFS count=1 × 11
+ * while libva did CREATE_BUFS count=24. If pool depth or buffer-index
+ * range affects rkvdec's per-codec state machine, this is a discriminator.
+ * Revert to 24 if 11 is too tight for mpv-16-surface continuous playback.
+ */
+#define MIN_CAP_POOL 11
 
 enum cap_slot_state {
 	CAP_SLOT_FREE = 0,	/* available for a new decode acquisition */