Prevent race when performing cache maintenance

[mezz]

Problem

There is a race condition, where maintenance can run after a cache hit is confirmed but before the files can be written.

Fix

To simplify the things I need to protect from races, I moved the "writing" logic out of RunNeoFormCommand#execute and into NeoFormEngine#writeResults, then I sealed off everything else in NeoFormEngine as private so that it only has that one public entry-point.

From there, I could lock the whole method with a read lock, and made it so that the cache maintenance needs to get a write lock so it cannot run when the NeoFormEngine is active.

ReadWriteFileLock

That just left one problem: we need a way for multiple threads and processes across multiple JVMs to have a read-write lock, and there's no re-entrant ReadWrite File Lock in Java.
Unfortunately (sorry) I created a new library to support this: https://github.com/mezz/ReadWriteFileLock
Please see the readme for justification for why I created this instead of using another library.
I wrote a test suite based on studying the other implementations, so it should cover all the necessary requirements for this project.

[neoforged-pr-publishing]

• Publish PR to GitHub Packages