LibWasm: Fix SIMD shuffle and swizzle

`swizzle` had the wrong operands, and the vector masking boolean logic
was incorrect in the internal `shuffle_or_0` implementation. `shuffle`
was previously implemented as a dynamic swizzle, when it uses an
immediate operand for lane indices in the spec.

AK/SIMDExtras.h

@@ -220,13 +220,14 @@ ALWAYS_INLINE static T shuffle_or_0_impl(T a, Control control, IndexSequence<Idx
     if constexpr (__has_builtin(__builtin_shuffle)) {
         // GCC does a very bad job at optimizing the masking, while not recognizing the shuffle idiom
         // So we jinx its __builtin_shuffle to work with out of bounds indices
+        // TODO: verify that this masking logic is correct (for machines with __builtin_shuffle)
         auto mask = (control >= 0) | (control < N);
         return __builtin_shuffle(a, control & mask) & ~mask;
     }
     // 1. Set all out of bounds values to ~0
     // Note: This is done so that  the optimization mentioned down below works
     // Note: Vector compares result in bitmasks, aka all 1s or all 0s per element
-    control |= ~((control > 0) | (control < N));
+    control |= ~((control >= 0) & (control < N));
     // 2. Selectively set out of bounds values to 0
     // Note: Clang successfully optimizes this to a few instructions on x86-ssse3, GCC does not
     //       Vector Optimizations/Instruction-Selection on ArmV8 seem to not be as powerful as of Clang18