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)
```

From: @yao_yf
Reviewed-by: @zhunaipan,@stsuteng
Signed-off-by: @stsuteng

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

Merge pull request !7437 from yao_yf/auto_parallel_support_dynamic_shape

Merge pull request !7495 from Xiaoda/27-star-elimination-overload

Merge pull request !4614 from zhoufeng/code-compile-performance

Merge pull request !4650 from yangzhenzhang/embedding-lookup-auto-parallel

Signed-off-by: zhoufeng <zhoufeng54@huawei.com>