Update qoa.h to latest git

thirdparty/README.md

@@ -792,7 +792,7 @@ Collection of single-file libraries used in Godot components.
     - `polypartition-0002-shadow-warning.patch` (GH-66808)
 - `qoa.{c,h}`
   * Upstream: https://github.com/phoboslab/qoa
-  * Version: git (a2d927f8ce78a85e903676a33e0f956e53b89f7d, 2024)
+  * Version: git (ae07b57deb98127a5b40916cb57775823d7437d2, 2025)
   * License: MIT
   * Modifications: Added implementation through `qoa.c`.
 - `r128.{c,h}`

thirdparty/misc/qoa.h

@@ -387,7 +387,7 @@ unsigned int qoa_encode_frame(const short *sample_data, qoa_desc *qoa, unsigned
 			int slice_start = sample_index * channels + c;
 			int slice_end = (sample_index + slice_len) * channels + c;
 
-			/* Brute for search for the best scalefactor. Just go through all
+			/* Brute force search for the best scalefactor. Just go through all
 			16 scalefactors, encode all samples for the current slice and
 			meassure the total squared error. */
 			qoa_uint64_t best_rank = -1;
@@ -402,7 +402,7 @@ unsigned int qoa_encode_frame(const short *sample_data, qoa_desc *qoa, unsigned
 				/* There is a strong correlation between the scalefactors of
 				neighboring slices. As an optimization, start testing
 				the best scalefactor of the previous slice first. */
-				int scalefactor = (sfi + prev_scalefactor[c]) % 16;
+				int scalefactor = (sfi + prev_scalefactor[c]) & (16 - 1);
 
 				/* We have to reset the LMS state to the last known good one
 				before trying each scalefactor, as each pass updates the LMS