[BATCH-2762] Introducing SynchronizedItemStreamWriter

[dimitrisli]

Also introducing SynchronizedItemStreamWriterBuilder. Multi-threaded
Tests have been that additionally prove that currently ItemStreamWriter
is not thread-safe.

[dimitrisli]

The multi-threaded tests are inspired by the construct in SynchronizedItemStreamReaderTests but it follows slightly different approach:

• using and atomic integer to track access to the write method
• a CountDownLatch to get all the threads ready
• assertion within the write method (closely paired with the atomic integer above)

Also realised that the existing SynchronizedItemStreamReaderTests is not actually properly working as the threads are not being started correctly - they are just invoking directly the run() method. I can fix this on a separate PR.

[fmbenhassine]

@dimitrisli Thank you for this PR! The synchronized item writer and its builder LGTM. However, when I see the tests, it feels like we are trying to test the synchronized modifier of Java. I would not bother testing that only one thread is calling the method at a time. I really trust the JVM here.

I think calling the synchronized writer and make sure the delegate was called is already ok. I know you said you took inspiration from SynchronizedItemStreamReaderTests, but those tests could be simplified as well. I will give you more details on that in your other PR #666.

I added a couple of minor inline comments, if you can address them the PR should be good to merge. Thank you upfront.

[fmbenhassine]

Thank you for the updates @dimitrisli !

However, when I see the tests, it feels like we are trying to test the synchronized modifier of Java. I would not bother testing that only one thread is calling the method at a time. I really trust the JVM here. I think calling the synchronized writer and make sure the delegate was called is already ok.

Do you agree on simplifying tests? I did not request a change for that (I can do it when merging your PR unless you want to push a second update), but at least I wanted to know if you agree on that.

[dimitrisli]

@benas totally agree. I've removed the concurrency tooling and focused on the delegate pass-through methods calling

[fmbenhassine]

@dimitrisli Great! LGTM now. Rebased, squashed and merged as 117b7d5. Please note that I applied some minor changes (update year in license headers, update javadoc, etc, see 9fbd3da).

Thank you for your contribution!