Fix shift-sign issue in Convert.c (python-pillow#7838)

* Fix shift-sign issue in Convert.c

Fixes
```
libImaging/Convert.c:513:25: error: signed shift result (0xFF000000) sets the sign bit of the shift expression's type ('int') and becomes negative [-Werror,-Wshift-sign-overflow]
    UINT32 trns = (0xff << 24) | ((b & 0xff) << 16) | ((g & 0xff) << 8) | (r & 0xff);
                   ~~~~ ^  ~~
```

---------

Co-authored-by: Andrew Murray <[email protected]>

Author: r-barnes

src/libImaging/Convert.c

@@ -510,7 +510,7 @@ rgbT2rgba(UINT8 *out, int xsize, int r, int g, int b) {
     UINT32 trns = ((r & 0xff) << 24) | ((g & 0xff) << 16) | ((b & 0xff) << 8) | 0xff;
     UINT32 repl = trns & 0xffffff00;
 #else
-    UINT32 trns = (0xff << 24) | ((b & 0xff) << 16) | ((g & 0xff) << 8) | (r & 0xff);
+    UINT32 trns = (0xffU << 24) | ((b & 0xff) << 16) | ((g & 0xff) << 8) | (r & 0xff);
     UINT32 repl = trns & 0x00ffffff;
 #endif