1. Provide method to create Parameter from Tensor without copy;
2. Adapt optimizer.py to support flattened parameters;
3. Add unit test cases for parameter and optimizer.

Merge pull request !23069 from huchunmei/master

Merge pull request !22509 from LiangZhibo/fix

- Fix the key generating.
- Distinguish the executors.

adjut the file structure and name

Deleted extra information

Do some formatting work

Add test case and fix some document

fix imports

This reverts commit b077aa1cab.

Revert "[feat] [assistant] [I3T96X] add new Dataset operator LibriSpeechDataset"

This reverts commit 4e6f7dc97d.

delete pass_registry_test.cc

comment  hiai_nlu_model_multi.pb related  line

split attention inputs from two into three

Add input check and modify test case

Fix check error and add check to transformer inputs

add input check to the transformers

Merge pull request !21351 from zhouyaqiang0/complex_support

---
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)
```

From: @zichun_ye
Reviewed-by: @wang_zi_dong,@sunnybeike
Signed-off-by:

fix typo in probability module

fix pylint in msp

From: @zyx5256
Reviewed-by: @liangchenghui,@kingxian
Signed-off-by: @liangchenghui