Fix the accelerate dispatch for the power op (#70)

- The exponent and base were swapped because accelerate is using
  exponent-base instead of base-exponent
- Fix also the test for binary ops as it was testing op(x, x) which
  couldn't catch ordering errors like that

mlx/backend/accelerate/primitives.cpp

@@ -494,7 +494,7 @@ void Power::eval_cpu(const std::vector<array>& inputs, array& out) {
       b.flags().row_contiguous) {
     int size = a.size();
     out.set_data(allocator::malloc_or_wait(out.nbytes()));
-    vvpowf(out.data<float>(), a.data<float>(), b.data<float>(), &size);
+    vvpowf(out.data<float>(), b.data<float>(), a.data<float>(), &size);
   } else {
     eval(inputs, out);
   }