[MTIA] Add MTIA dispatch for kernel foreach_maximum(Add D80022242 back) (pytorch#161571)

Summary: dispatch MTIA to function foreach_tensor_maximum_scalar_kernel_mtia_

Test Plan:
CI

Rollback Plan:

Differential Revision: D81086607

Pull Request resolved: pytorch#161571
Approved by: https://github.com/malfet

Author: DoubleBiao

aten/src/ATen/native/native_functions.yaml

@@ -10848,6 +10848,7 @@
   dispatch:
     CompositeExplicitAutograd: foreach_tensor_clamp_min_scalar_kernel_slow_
     CUDA: foreach_tensor_clamp_min_scalar_kernel_cuda_
+    MTIA: foreach_tensor_maximum_scalar_kernel_mtia_
   autogen: _foreach_maximum.Scalar_out
 
 # foreach_minimum/maximum dispatches to clamp_max/min