[roll] Roll fuchsia [runtime] Optimize extra atomic acquire load in _zx_thread_self

libsync and pthread mutexes call _zx_thread_self() when acquiring a lock
to retrieve a 32 bit thread identity value with the low bit clear. This
function retrieves the handle value at an offset from the thread
pointer using zxr_thread_get_handle() to do the lookup. This function
can also be used to retrieve the handle value for a different thread and
as such a synchronizing atomic acquire load was introduced in commit
1db2dc84f6f24fee118f750e8ea8aa2278d9e5de to make these cases safe. This
load is not needed when retrieving the calling thread's handle value.

This adds a zxr routine to be used only for the calling thread's value
and uses that in _zx_thread_self(). This removes an atomic acquire load
from the mutex acquisition path.

Original-Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/1207605
Original-Revision: 46a89539bab86e6f47cd5e2f3826d306df038ebc
GitOrigin-RevId: b0ee35a952a2ad48196f87f9edb187c7e107d2ef
Change-Id: I44b6124aabbec24fabdec8e79cdb3001ea868322