*: log each com_stmt_fetch separately

[lysu]

What problem does this PR solve?

to slow log:

In the current master code, slow log will be wrong value(last fetch time(or new start execute time) - last cmd time)
In 2.1 code, slow log will be incompatible value(last fetch time(or new start execute time) - first com_stmt_execute time)
but mysql record com_stmt_exec and com_stmt_fetch for each request.

this pr will change the behavior as MySQL does, although TiDB and MySQL use very different way to handle cursor:

• MySQL execute sql and save the result in temp table(memory or myisam), then fetch the tmp table
• TiDB cache the executor-tree and execute when fetch command came and cache partial result in memory(now TiDB-server node is stateless and doesn't write any data in local disk)

so, you may see mysql stmt_execute is slow but fetch super fast, and TiDB both execute and fetch is slow in log.

What is changed and how it works?

• add LogParitialSlow to recordset
• Fix stmt_reset should clean up opened cursor in stmt(but pay attention both libmysql c API and jdbc, close ResultSet shouldn't send stmt_reset in current client impl, but mysql server will handle stmt_reset like this)
• log InPreapre and HasMoreResult in slow log too(help find question)
• log Parse_time and Compile_time in slow log(it seems miss those two fields)

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Code changes

• Has exported function/method change
• Has exported variable/fields change
• Has interface methods change
• Has persistent data change

Side effects

• Possible performance regression
• Increased code complexity
• Breaking backward compatibility

Related changes

• Need to cherry-pick to the release branch
• Need to update the documentation
• Need to update the tidb-ansible repository

Release note

• Write release note for bug-fix or new feature.

---

This change is 

[lysu]

/run-all-tests

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@4891e18). Click here to learn what that means.
The diff coverage is 93.75%.

@@            Coverage Diff             @@
##             master    #11987   +/-   ##
==========================================
  Coverage          ?   80.745%           
==========================================
  Files             ?       454           
  Lines             ?     99403           
  Branches          ?         0           
==========================================
  Hits              ?     80263           
  Misses            ?     13346           
  Partials          ?      5794

[lonng]

IMO, I don't like this way to fix the slow query misinformation (the extra LogPartialSlow method is not an elegant way). Prefer to change the query cost time to an accumulative time and accumulate it every com_stmt_fetch.

[jackysp]

LGTM

[lysu]

IMO, I don't like this way to fix the slow query misinformation (the extra LogPartialSlow method is not an elegant way). Prefer to change the query cost time to an accumulative time and accumulate it every com_stmt_fetch.

mysql record slow log for each comand time, should keep compatible with that?

https://github.com/mysql/mysql-server/blob/2f553f258daa213c710bc56d4fd7b8faeb7c4631/sql/sql_parse.cc#L1458

https://github.com/mysql/mysql-server/blob/2f553f258daa213c710bc56d4fd7b8faeb7c4631/sql/sql_parse.cc#L1738

[jackysp]

PTAL @lonng .
Please resolve the conflicts, @lysu .

[coocood]

The method name is too specific.
Something like OnFetchDone would be better.

[lysu]

I tried remove this method out of that interface and give a new name now~ PTAL if free

[lysu]

/run-all-tests

[lysu]

@coocood @tiancaiamao PTAL

[coocood]

LGTM

[jackysp]

LGTM

[sre-bot]

/run-all-tests

[sre-bot]

cherry pick to release-3.1 failed

[sre-bot]

cherry pick to release-3.0 in PR #12392