Reduce system calls of ThreadPriorityBooster

- ThreadLocal is bound to each thread, so we don't need to
  get tid every time.
- By storing whether the priority is changed, we don't need
  to get thread priority to decide whether to restore.

The cost of a simple boost region is reduced over 20%.

Test: atest ThreadPriorityBoosterTest
Bug: 123043091
Change-Id: I634898c1bf0d10de4f53fd18043f64d18fe9082b