[INLONG-7249][Sort] JDBC accurate dirty data archive and metric calculation

[Yizhou-Yang]

the pr needs to remove logs and add comments before the final merge

Fixes #7249

Feature: record level accurate dirty data archive and metric for jdbc connector

Design:

1. Use java reflection to override TableSimpleStatementExecutor, pass in inlong metric and dirty helper for accurate metrics.
2. Add a counter, to not do record level flush for the first 3 times in case of network problems
3. Parse out the error record from error message
4. Add batch mechanism and iterate the batch to find the position of dirty data
5. Refresh batch, update metrics, and retry the records after dirty data.

[Yizhou-Yang]

todo:
for multiple sink, avoid redundent metric calculation when the executor is enhanced
check compatibility and performance in various scenerios

some thoughts on this pr:
1.this pr is DEPENDENT on the variable naming of specific flink executors. Hopefully the namings in flink won't be changed.
2.the long[] for metrics is not thread safe on write. I think this is ok because flush only happens after a certain interval/batch size, and there should not be concurrent modification.

[Yizhou-Yang]

TODO : the pr is still being performance-tested and need to remove logs and add comments before the final merge

[Yizhou-Yang]

I will reorganize the code and produce a new pr sometime later this month