From: @zhangzhaoju
Reviewed-by: @zh_qh,@ginfung
Signed-off-by: @zh_qh

2. Dump the end graph in gpu session and cu session

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

---
Remove the routine of handling isolated nodes in Renormalize.
Add isolated nodes from Parser&Resolver.
Modify isolated nodes handling in FG&Manager.
Optimize the renormalize routines.
Other optimizations.

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)
```

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

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

Merge pull request !2948 from amongo/FixControlFlow

add join for ref

adjust env eliminate to eliminate all env ops

add partial app cache

resolve while endless

fix env eliminate

support for "for while" cases

fix join shape error

Merge pull request !2017 from leopz/fix_codedex

Signed-off-by: leonwanghui <leon.wanghui@huawei.com>

Signed-off-by: leonwanghui <leon.wanghui@huawei.com>