[RFC] ntdll: improve heap allocation performance

[Nils Kuhnhenn]

I've been looking into why, in GW2, HEAP_FindFreeBlock was using up to
20-40% of the total CPU time consumed by the application if GW2 ran
for a longer timer and/or when there is a lot going on on-screen.

The result is this patch which aims to speed up memory allocation performance
for applications that have bad behavior, but should affect applications
that behave nicely very little, if at all.

For me the result is that GW2 now spends less than 2% CPU time in HEAP_*
functions and doesn't drop to an unplayable FPS rate in WvW/after
a few hours.

More information is in the commit message.