The basic idea is: exploits data dependency to control the execution order
of side-effect operations, and keep the semantics of ANF unchanged.

The ControlDepend primitive is removed and there are two primitives added:

1. UpdateState:
```
  a = Assign(para, value)
```
became:
```
  a = Assign(para, value, u)
  u = UpdateState(u, a)
```

2. Load:
```
  x = Add(para, value)
```
became:
```
  p = Load(para, u)
  x = Add(p, value)
  u = UpdateState(u, p)
```

Fix can not collect input data when batch size is 1 and total step
number is 1

Fixed spelling errors

From: @gzhcv
Reviewed-by: 
Signed-off-by:

Merge pull request !7855 from huanghui/move_ir_files

Signed-off-by: zjun <zhangjun0@huawei.com>

Merge pull request !7373 from lilei/modify_MetaTensor

Merge pull request !5735 from caozhou/delete_seed

Merge pull request !5885 from vlne-v1/add-doc-for-composite-op

clear df builder

add testcases

Merge pull request !5669 from hewei/fix_get_py_obj_dtype

Merge pull request !5502 from wanyiming/mod_SoftmaxCrossEntropyWithlogits

Return mstype.type_type if input is mindspore type object,
for example:
```
get_py_obj_dtype(mstype.float32) ---> mstype.type_type
```

Merge pull request !4922 from amongo/TupleTransform

add support for pynative pass

add testcases

Merge pull request !5026 from Simson/enhancement-API

Merge pull request !5011 from riemann_penn/remove_global_grad_ops

add testcase of net of if by if

Merge pull request !4370 from zhangbuxue/support_kw_and_kwargs_for_cell_in_pynative