Reduce default size of diagnostic buffer for FlightRecorderInputStream from 1MiB to 1KiB

[dwnusbaum]

One of the main side effects of the bug fixed by jenkinsci/google-compute-engine-plugin#478 was that each failed connection caused a 1MiB buffer to be retained by a FlightRecorderInputStream, leading to high memory usage. This buffer is only used to print a hex dump of the stream contents in case of a stream corruption error, and so having it be a full 1MiB by default seems unnecessary, and if there is something interesting in the dump, it is probably right at the end.

I considered disabling the buffer completely by making the default value 0, but that requires a few changes to avoid ArrayIndexOutOfBoundsException and improve the error messages, so it seemed more expedient to just reduce the default buffer size.

In the real-world case we observed, there were around 600 instances of this class in memory, so this patch would have cut ~600MB of memory used by these buffers down to around ~600KB.

See jenkinsci/jenkins#9887 for the matching core PR.

Testing done

Only tested using the existing automated tests in FlightRecorderInputStreamTest and DiagnosedStreamCorruptionExceptionTest.

Submitter checklist

• Make sure you are opening from a topic/feature/bugfix branch (right side) and not your main branch!
• Ensure that the pull request title represents the desired changelog entry
• Please describe what you did
• Link to relevant issues in GitHub or Jira
• Link to relevant pull requests, esp. upstream and downstream changes
• Ensure you have provided tests - that demonstrates feature works or fixes the issue

[dwnusbaum]

If you think this is too aggressive, we could increase to 10KiB.