Add BulkByScrollResponse Serialization Tests

[joshua-adams-1]

Adds BulkByScrollResponseWireSerializingTests and removes a redundant testRoundTrip() test from BulkByScrollResponseTests

[joshua-adams-1]

There is greater serialisation test coverage from BulkByScrollResponseserializationTests than this test previously provided

[elasticsearchmachine]

Pinging @elastic/es-distributed (Team:Distributed)

[PeteGillinElastic]

I confess that I'm a little doubtful about the utility of this test. It seems to me that the most common cause of regressions in the wire serialization code is someone adding a new field and forgetting to update the serialization. But if they've forgotten that, they're definitely not going to remember to update this test, which will continue to pass.

The only alternative to that would be to do something reflection-based, which seems overkill (and is, I assume, not something that we're doing elsewhere).

Still, it's probably going to do more good than harm, so LGTM.

[joshua-adams-1]

I confess that I'm a little doubtful about the utility of this test. It seems to me that the most common cause of regressions in the wire serialization code is someone adding a new field and forgetting to update the serialization. But if they've forgotten that, they're definitely not going to remember to update this test, which will continue to pass.

I think this would only be true if someone added a new field but didn't change the writeTo method, although in that case we haven't technically changed the serialisation, and so this test should still be passing, right?

What this test will catch in future is:

• Writing a field without a transport‑version guard
• Reading a field unconditionally
• Gating on the wrong version
• Changing field order across versions
• A field is moved, renamed, or type‑changed

etc, which is much better coverage than before. Plus it tests equality and hashcode, which the original roundTrip() method did not