[CFL] Change CFL DC_PRED from int to int16_t

This changes serves to illustrate that DC_PRED does not
need 32 bits. It only needs the specified pixel bit depth.
Which fits inside a signed 16 bit integer.

Change-Id: Ice062c8c4a794cb4f32a3963278f4a99ef326dda

av1/common/blockd.h

@@ -604,7 +604,8 @@ typedef struct cfl_ctx {
   int subsampling_x, subsampling_y;
 
   // Block level DC_PRED for each chromatic plane
-  int dc_pred[CFL_PRED_PLANES];
+  // 16 bits are used, but only the pixel bit depth is required
+  int16_t dc_pred[CFL_PRED_PLANES];
 
   int mi_row, mi_col;
 

av1/common/cfl.c

@@ -259,7 +259,7 @@ static INLINE int cfl_idx_to_alpha(int alpha_idx, int joint_sign,
 
 static void cfl_build_prediction_lbd(const int16_t *pred_buf_q3, uint8_t *dst,
                                      int dst_stride, int width, int height,
-                                     int alpha_q3, int dc_pred) {
+                                     int alpha_q3, int16_t dc_pred) {
   for (int j = 0; j < height; j++) {
     for (int i = 0; i < width; i++) {
       dst[i] =
@@ -273,7 +273,8 @@ static void cfl_build_prediction_lbd(const int16_t *pred_buf_q3, uint8_t *dst,
 #if CONFIG_HIGHBITDEPTH
 static void cfl_build_prediction_hbd(const int16_t *pred_buf_q3, uint16_t *dst,
                                      int dst_stride, int width, int height,
-                                     int alpha_q3, int dc_pred, int bit_depth) {
+                                     int alpha_q3, int16_t dc_pred,
+                                     int bit_depth) {
   for (int j = 0; j < height; j++) {
     for (int i = 0; i < width; i++) {
       dst[i] = clip_pixel_highbd(