Add test for halide#8029 (halide#8032)

Tweak correctness_float16_t so that it uses one of the transcendal functions (sqrt) that were missing in Metal.

test/correctness/float16_t.cpp

@@ -236,14 +236,14 @@ int run_test() {
  Param<float16_t> mul("mul");
 
  Func output;
- output(x, y) = x * y * (input(x, y) * mul);
+ output(x, y) = x * y * (sqrt(input(x, y)) * mul);
 
  Var xi, yi;
  output.gpu_tile(x, y, xi, yi, 8, 8);
 
  mul.set(float16_t(2.0f));
  Buffer<float16_t> in(8, 8);
- in.fill(float16_t(0.25f));
+ in.fill(float16_t(0.0625f));
  input.set(in);
  Buffer<float16_t> buf = output.realize({8, 8});
  for (int y = 0; y < 8; y++) {