v4.6.2

Changed the language server startup process to remove all files from memory after initial startup. This reduces RAM by 0.25-0.5 GB of memory after startup, depending on the volume of code in workspaces and on your path.

When we clean up the language server (happens every 5 minutes), we now check our in-memory cache and remove any files that haven't been accessed recently. Helps reduce overall RAM when VSCode is open for long periods at a time.

Changed the way that the language server sends work to the threads that parse code. The main difference is that we no longer send the parsed version of code back to the main thread which we were doing.

For large files, this had a significant impact on perceived performance as the worker threads could get locked serializing and de-serializing objects (also leads to more memory usage). Now, large files like slicer3.pro which are included in the IDL installation, provide auto-complete and hover help in about ~0.5 seconds instead of 5+ seconds.

Fixed an issue where, from the JSON settings editor, a valid extension setting would be highlighted as being an incorrect value.

Fixed an issue where the "don't ask again" options didn't honor your choice for using IDL as the formatter.