refactor: Restructure SQLRecognizer and UndoExecutor

[booogu]

<! -- make sure you have read and understood the contribution guide -- >

###I. describe what this PR has done
On the side of RM,

1. Add UndoExecutorGroupand and its corresponding factory. The UndoExecutorGroup interface abstracts executor that all kinds of databases need to provide (currently implemented by MySQL and Oracle)

2. Add SqlRecognizerGroup and its corresponding factory. SqlRecognizerGroup abstracts the recognizer (currently mqsql and Oracle) that all kinds of databases need to provide,

The benefits of this reconstruction are as follows:

1. Eliminate redundant code.

2. Enhance the scalability of the project (a new type of database can be added on RM side through resource configuration file)

3. When a new database type is to be added in the future, the capabilities it needs to provide will be roughly divided into three dimensions:

(1) extend AbstractUndoLogManager. (An existing abstraction provided in a previous PR,)

(2) implement UndoExecutorGroup.

(3) implement SQLRecognizerGroup.

To sum up, after refactoring, For the provider (the adaptation of each database type), will be defined and abstracted more clearly and cleanly; For the user (where to get the instance of executor or recognizer), it will be more convenient by calling factory methods directly.

###Does this request solve a problem?

<! -- if so, add "fixesාxxx" to the next line, for example, fixes񖓿97.

###III. why not add test cases (unit test / integration test)?

This PR is only for the adjustment of code structure, not involving the change of specific logic.

###IV. describe how to verify

It only adjusts the internal abstraction level and calling mode, without any specific logical changes. After pulling the latest code, test the official Seata demo project, and conduct overall verification.

###V. special notes for review

Note that in this submission, the two factories I provided (undoexecutor groupfactory and sqlrecognizergroupfactory) are the instances created directly by using the EnhancedServiceLoader to load specific implementation classes from the resource file,

This is mainly to consider: Although there are only two kinds of database adaptations now: mqsql and Oracle, in the future, when other developer expand other types of databases, they can directly add resource file to provide their implemention without changing the existing code.

[codecov-io]

Codecov Report

Merging #1883 into develop will increase coverage by <.01%.
The diff coverage is 82.5%.

@@             Coverage Diff              @@
##             develop   #1883      +/-   ##
============================================
+ Coverage       55.3%   55.3%   +<.01%     
- Complexity      2388    2408      +20     
============================================
  Files            422     428       +6     
  Lines          14300   14318      +18     
  Branches        1702    1659      -43     
============================================
+ Hits            7908    7918      +10     
+ Misses          5678    5672       -6     
- Partials         714     728      +14

Impacted Files	Coverage Δ	Complexity Δ
...source/sql/druid/MySqlOperateRecognizerHolder.java	100% <100%> (ø)	6 <6> (?)
...datasource/undo/mysql/MySQLUndoExecutorHolder.java	100% <100%> (ø)	5 <5> (?)
...ql/druid/oracle/OracleOperateRecognizerHolder.java	100% <100%> (ø)	6 <6> (?)
...tasource/undo/oracle/OracleUndoExecutorHolder.java	50% <50%> (ø)	2 <2> (?)
.../seata/rm/datasource/undo/UndoExecutorFactory.java	64.28% <69.23%> (+42.85%)	3 <3> (ø)	⬇️
.../rm/datasource/undo/UndoExecutorHolderFactory.java	73.33% <73.33%> (ø)	3 <3> (?)
...asource/sql/SQLOperateRecognizerHolderFactory.java	86.66% <86.66%> (ø)	4 <4> (?)
.../io/seata/rm/datasource/sql/SQLVisitorFactory.java	83.33% <90.9%> (-2.39%)	6 <0> (-6)
...obuf/convertor/BranchRegisterRequestConvertor.java	88% <0%> (-8%)	3% <0%> (ø)
...buf/convertor/GlobalLockQueryRequestConvertor.java	92.59% <0%> (-7.41%)	3% <0%> (ø)
... and 32 more

[zjinlei]

Thanks for submitting again.
I am sorry to have closed Https://github.com/seata/seata/pull/1523

[booogu]

Thanks for submitting again.
I am sorry to have closed Https://github.com/seata/seata/pull/1523

It doesn't matter. That time, I accidentally messed up the code format locally, making the difference not easy to compare. Moreover, my change was later included in the PR of other developers.

[zjinlei]

@ggndnn please review together

[zjinlei]

@CharmingRabbit Please link your git account email when submitting
refer: https://blog.csdn.net/idealcitier/article/details/82697268

[booogu]

@CharmingRabbit Please link your git account email when submitting
refer: https://blog.csdn.net/idealcitier/article/details/82697268

Thanks for the link! I've set it up

[zjinlei]

please resolve CI error
https://travis-ci.org/seata/seata/jobs/610220930?utm_medium=notification&utm_source=github_status

[slievrly]

@CharmingRabbit please move all @date element, later we will also remove the existing ones, here you are datetime instead of date. @dateis not a standard javadoc element.

[booogu]

please resolve CI error
https://travis-ci.org/seata/seata/jobs/610220930?utm_medium=notification&utm_source=github_status

This Ci error is caused by the rename of the class. I will solve it later

[booogu]

@CharmingRabbit please move all @date element, later we will also remove the existing ones, here you are datetime instead of date. @dateis not a standard javadoc element.

removed

[slievrly]

LGTM

[zjinlei]

retry ci

[booogu]

retry ci

passed

[zjinlei]

LGTM

[zjinlei]

@CharmingRabbit Thank you for your contribution.
There are 2 minor problems, and the latter will be rectified.
1.NPE: !dbType.equals(JdbcConstants.MYSQL) && !dbType.equals(JdbcConstants.ORACLE)
2.code format

[ggndnn]

@ggndnn please review together

Sorry, didn't notice this comment in time.

[booogu]

@CharmingRabbit Thank you for your contribution.
There are 2 minor problems, and the latter will be rectified.
1.NPE: !dbType.equals(JdbcConstants.MYSQL) && !dbType.equals(JdbcConstants.ORACLE)
2.code format
OK，got it.
finally，@zjinlei @slievrly thank you very much for your help in my first code contribution.