[MPS] Fix sqrt and other for torch.chalf (pytorch#148285)

Those kernels, instead of being instantiated for half2 (which corresponds to ComplexHalf) were instnatiated for short2, which resuled in the following test
```
% python3 -c "import torch; print(torch.rand(6, device='mps', dtype=torch.chalf).sqrt())"
```
Fail with
```
RuntimeError: Failed to create function state object for: sqrt_complex_half_half
```

As sqrt is not implemented for CPU, add explicit test to `test_sqrt`
Pull Request resolved: pytorch#148285
Approved by: https://github.com/dcci

Author: malfet

aten/src/ATen/native/mps/kernels/UnaryKernel.metal

@@ -146,7 +146,7 @@ INSTANTIATE_UNARY_KERNELS2(float, long);
       constant vec2type_t<DTYPE0> * input [[buffer(1)]],                  \
       uint did [[thread_position_in_grid]]);
 
-INSTANTIATE_UNARY_KERNELS_VEC2(short, short);
+INSTANTIATE_UNARY_KERNELS_VEC2(half, half);
 INSTANTIATE_UNARY_KERNELS_VEC2(float, float);
 
 template <typename T0, typename T1>

test/test_mps.py

@@ -6854,6 +6854,12 @@ def helper(shape):
 
         helper((2, 8, 4, 5))
 
+        # Test complex half
+        x = torch.rand(8, device='mps', dtype=torch.chalf)
+        rc_h = x.sqrt()
+        rc_f = x.cfloat().sqrt().chalf()
+        self.assertEqual(rc_h, rc_f)
+
     # Test selu, elu, celu
     def test_elu(self):
         def helper(shape, alpha=1.0, memory_format=torch.contiguous_format):