perf: schedule in eventloop for batchWrite of endpoint

[okg-cxf]

Make sure that:

• [Y] You have read the contribution guidelines.
• [Y] You have created a feature request first to discuss your contribution intent. Please reference the feature request ticket number in the pull request.
• [Y] You applied code formatting rules using the mvn formatter:format target. Don’t submit any formatting related changes.
• [Y] You submit test cases (unit or integration tests) that back your changes.

channelWrite() will result in SingleThreadEventLoop#taskQueue.offer() which is a volatile write, it could be quite expensive if the batch size is super large.

Bench result (batchSize: 500)

with perf:
RedisClientBenchmark.asyncSetBatchFlush     avgt   10   1176.188 ±    2.945  ns/op  (18% performance gain)
RedisClientBenchmark.reactiveSetBatchFlush  avgt   10   7701.130 ±  580.995  ns/op (performance slightly down, maybe need to add a switch to control the behavior?)

without perf:
Benchmark                                   Mode  Cnt      Score      Error  Units
RedisClientBenchmark.asyncSetBatchFlush     avgt   10   1387.984 ±   46.847  ns/op
RedisClientBenchmark.reactiveSetBatchFlush  avgt   10   7308.346 ±  129.363  ns/op

[tishun]

The failing tests are failing in other PRs too, seems to be a regression, investigating currently ...

[tishun]

The failing tests are failing in other PRs too, seems to be a regression, investigating currently ...

You need to fetch the fix from #2949

[Roiocam]

Looks like it will limit the throughput because it will has only few thread write, in the legacy way, both IOGroup and ExecutorGroup can execute.

can you verify the performance improvement not just JMH case?