Improve startup performance

PDE startup is slow because much of the initialization work happens on the main thread, blocking the application from becoming responsive.

PR #1300 explored moving several initialization tasks to background threads and demonstrated that this can significantly improve startup performance. The PR was ultimately closed in favor of revisiting startup concurrency with a more centralized approach, rather than introducing multiple ad-hoc threads in startup code.

Relevant quotes:

@tychedelia: “Have you considered using an ExecutorService instead of manually creating all the threads? That could provide better control or allow thread reuse particularly for tasks that are small.”

@catilac: “If there is a thread pool manager we can use to manage threads we should use that. I'd prefer we not introduce the use of low-level concurrency primitives on the app startup code paths.”

I'm opening this issue to capture these insights, so startup performance work can be revisited later.

Start by reviewing PR #1300 and the quoted discussion about ExecutorService and a shared thread-pool manager. Identify the centralized startup-concurrency approach to pursue and define which initialization tasks can move off the main thread. Done means startup becomes responsive without introducing ad-hoc threads, with performance evidence for the change.