Optimize Thread Wait Logic in BlockableEventLoop

[HaHaWTH]

This PR should fix MC-183518. The current implementation in BlockableEventLoop involves yielding the thread before parking, which is unnecessary and can lead to inefficient thread scheduling and higher CPU usage.

Changes made:

• Replaced the existing thread yielding logic with LockSupport.parkNanos to directly park the thread for 2 milliseconds.

Reasoning: Yielding the thread is redundant if the thread is about to be parked immediately afterward.

Reference:

• MC-183518

[Dreeam-qwq]

GO