Remove RetriableStream flush workaround

[ejona86]

In #9360 I added this workaround:

grpc-java/core/src/main/java/io/grpc/internal/RetriableStream.java

Lines 565 to 568 in 7c73baa

	// TODO(ejona): Workaround Netty memory leak. Message writes always need to be followed by
	// flushes (or half close), but retry appears to have a code path that the flushes may
	// not happen. The code needs to be fixed and this removed. See #9340.
	substream.stream.flush();

A flush() must always follow a write (but other flushes and calls could be in-between). I tried to flush everywhere but it didn't work at the time. See also #9340. It is possible there were interactions with #9547 which prevented it from working and just trying again may work.

[AppleTreeQ]

Hi, I am new to grpc-java.
I would like to make contributions to this project.
Would you like assign this task to me. Or would you like to assign me a good-first-issue?
Thank you very much.

[ejona86]

@AppleTreeQ, I know of someone else who was also probably going to look into this. But he'd be a bit more experienced with gRPC. This is a pretty poor first issue because it was flaky and the retry code has complicated synchronization.

We have a 'good first issue' label that you can glance at, but there are only a few items.

[AppleTreeQ]

@ejona86
Thank you for your reply.
I checked the 'good first issue' list, maybe I can work on #9643 if no one look into it.