avutil/bswap: fix implicit conversion warning in av_bswap64

Explicitly cast uint64_t arguments to uint32_t before passing them
to av_bswap32(). The truncation is intentional (extracting low and
high halves), but clang on macOS 26 warns about it.

Fixes: https://code.ffmpeg.org/FFmpeg/FFmpeg/issues/22453
Signed-off-by: marcos ashton <marcosashiglesias@gmail.com>

libavutil/bswap.h

@@ -68,7 +68,7 @@ static av_always_inline av_const uint32_t av_bswap32(uint32_t x)
 #ifndef av_bswap64
 static inline uint64_t av_const av_bswap64(uint64_t x)
 {
-    return (uint64_t)av_bswap32(x) << 32 | av_bswap32(x >> 32);
+    return (uint64_t)av_bswap32((uint32_t)x) << 32 | av_bswap32((uint32_t)(x >> 32));
 }
 #endif