Reduce prec of matrices/vectors for warp estimate

Reduces precision of matrices by 2 bits.

No material change in performance.

Change-Id: I549a27da1dcb381fb329c345ee280dbd86b45bac
diff --git a/av1/common/warped_motion.h b/av1/common/warped_motion.h
index 0e88619..dea7cfd 100644
--- a/av1/common/warped_motion.h
+++ b/av1/common/warped_motion.h
@@ -26,7 +26,10 @@
 #define MAX_PARAMDIM 9
 #if CONFIG_WARPED_MOTION
 #define SAMPLES_ARRAY_SIZE ((2 * MAX_MIB_SIZE + 2) * 2)
-#define LEAST_SQUARES_SAMPLES_MAX 8
+
+#define LEAST_SQUARES_SAMPLES_MAX_BITS 3
+#define LEAST_SQUARES_SAMPLES_MAX (1 << LEAST_SQUARES_SAMPLES_MAX_BITS)
+
 #define DEFAULT_WMTYPE AFFINE
 #endif  // CONFIG_WARPED_MOTION