bugfix: eliminate the impact of instructions reordering on session async committing task

[ggndnn]

Ⅰ. Describe what this PR did

After code analysis, I found there may be instructions reordering in DefaultCore#asyncCommit. Session object may be added to async committing session manager first, before globalSession.addSessionLifecycleListener and globalSession.changeStatus. The reason is that GlobalSession#lifecycleListeners and DefaultSessionManager#sessionMap have no dependency to each other.

    private void asyncCommit(GlobalSession globalSession) throws TransactionException {
        globalSession.addSessionLifecycleListener(SessionHolder.getAsyncCommittingSessionManager());
        SessionHolder.getAsyncCommittingSessionManager().addGlobalSession(globalSession);
        globalSession.changeStatus(GlobalStatus.AsyncCommitting);
    }

I think, unexpected scheduling delay of async committing job and instructions reordering lead to the situation happened in #1564.

The analysis of one error log:

[INFO] Running io.seata.server.event.DefaultCoreForEventBusTest
2019-08-30 09:34:45.530 INFO [UndoLogDelete_1]io.seata.server.coordinator.DefaultCoordinator.undoLogDelete:403 -no active rm channels to delete undo log
2019-08-30 09:34:45.629 INFO [AsyncCommitting_1]io.seata.server.coordinator.DefaultCore.doGlobalCommit:238 -Global[172.17.0.1:0:4020856891] committing is successfully done.
2019-08-30 09:34:45.630 INFO [AsyncCommitting_1]io.seata.server.coordinator.DefaultCore.doGlobalCommit:238 -Global[172.17.0.1:0:4020856891] committing is successfully done.
[ERROR] Tests run: 1, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 1.035 s <<< FAILURE! - in io.seata.server.event.DefaultCoreForEventBusTest
[ERROR] test  Time elapsed: 1.034 s  <<< FAILURE!
org.opentest4j.AssertionFailedError: expected: <1> but was: <2>
	at io.seata.server.event.DefaultCoreForEventBusTest.test(DefaultCoreForEventBusTest.java:82)

• First run of async committing job delayed to 09:34:45.629, session status was not changed to AsyncCommitting until after DefaultCore#doGlobalCommit#endCommitted, session object still be in session manager. One AsyncCommitting event was posted here.
• Second run of async committing job started as scheduled at 09:34:45.630, another AsyncCommitting event was posted then.

This PR only add a session status check, any better idea?

Ⅱ. Does this pull request fix one issue?

Ⅲ. Why don't you add test cases (unit test/integration test)?

Ⅳ. Describe how to verify it

Ⅴ. Special notes for reviews

[codecov-io]

Codecov Report

Merging #1576 into develop will increase coverage by 0.17%.
The diff coverage is 0%.

@@              Coverage Diff              @@
##             develop    #1576      +/-   ##
=============================================
+ Coverage      47.08%   47.25%   +0.17%     
  Complexity      1764     1764              
=============================================
  Files            359      359              
  Lines          12971    12999      +28     
  Branches        1579     1617      +38     
=============================================
+ Hits            6107     6143      +36     
- Misses          6187     6191       +4     
+ Partials         677      665      -12

Impacted Files	Coverage Δ	Complexity Δ
...o/seata/server/coordinator/DefaultCoordinator.java	50% <0%> (-0.46%)	26 <0> (ø)
...in/java/io/seata/server/session/GlobalSession.java	84.54% <0%> (-0.61%)	67% <0%> (ø)
...a/io/seata/core/rpc/netty/AbstractRpcRemoting.java	9.95% <0%> (-0.21%)	3% <0%> (ø)
...n/java/io/seata/rm/datasource/ConnectionProxy.java	22.88% <0%> (-0.2%)	10% <0%> (ø)
...tobuf/convertor/BranchCommitResponseConvertor.java	97.43% <0%> (+0.06%)	3% <0%> (ø)	⬇️
...in/java/io/seata/server/session/BranchSession.java	78.19% <0%> (+0.07%)	41% <0%> (ø)	⬇️
...buf/convertor/BranchRollbackResponseConvertor.java	97.36% <0%> (+0.07%)	3% <0%> (ø)	⬇️
...tobuf/convertor/GlobalCommitResponseConvertor.java	97.05% <0%> (+0.08%)	3% <0%> (ø)	⬇️
...buf/convertor/GlobalRollbackResponseConvertor.java	97.05% <0%> (+0.08%)	3% <0%> (ø)	⬇️
...tobuf/convertor/GlobalStatusResponseConvertor.java	97.05% <0%> (+0.08%)	3% <0%> (ø)	⬇️
... and 17 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9877828...4f1ead9. Read the comment docs.

[xingfudeshi]

LGTM.

[zjinlei]

LGTM