[CANN]Support operator SIN COS ARGMAX

[noemotiovon]

Why is this PR needed?

Optimize the sin , cos, argmax operator in the CANN backend with the aclnn acceleration library.

Test

  # SIN
  SIN(type=f16,ne=[10,2,2,2]): OK
  SIN(type=f32,ne=[10,2,2,2]): OK
  5294/5294 tests passed
  Backend CANN0: OK
  
  # COS
  COS(type=f16,ne=[10,2,2,2]): OK
  COS(type=f32,ne=[10,2,2,2]): OK
  5294/5294 tests passed
  Backend CANN0: OK
  
  #ARGMAX
  ARGMAX(type=f32,ne=[32,1,1,1]): OK
  ARGMAX(type=f32,ne=[100,10,1,1]): OK
  ARGMAX(type=f32,ne=[1024,10,1,1]): OK
  ARGMAX(type=f32,ne=[1024,12,1,1]): OK
  ARGMAX(type=f32,ne=[2000,10,1,1]): OK
  ARGMAX(type=f32,ne=[5438,3,1,1]): OK
  5294/5294 tests passed
  Backend CANN0: OK

[noemotiovon]

The extra copy here is necessary because the shape computed by the aclnn operator differs from dst, so a buffer is used to hold the data before copying it back to dst.

[hipudding]

No need to alloc a extra buffer. use dst->data instead.

[noemotiovon]

Thank you for your suggestion! I forgot to directly modify dst->data's shape, haha.

[hipudding]

It seems aclnn_argmax need dst, and acl_dst can be created in aclnn_argmax, and do not need pass as a parameter.
Besides, If aclnn_argmax is not used by other internal functions, these two function should combine together.

[noemotiovon]

Okay, I will inline this method.