Fixes an issue where a thread that exits with the `exit` syscall never
actually frees its pivot stack. This is common practice and it was
missed when I was fixing the previous stack pivot leak.
This was uncovered when looking at the game
[RUINER](https://store.steampowered.com/agecheck/app/464060/?curator_clanid=4777282)
for timing bugs. Turns out the Linux build of the game creates and
destroys a VLC object every tick of the engine. Creating this VLC
object creates six threads behind the scenes. At what I assume the
default tick of the engine is of 120Hz(?) this would mean it is
attempting to create and destroy 720 threads per second, quickly
leading to memory exhaustion under FEX.
While this hits the biggest memory leak we have around thread creation,
this game is still hitting thread creation so hard that I can see other
leaks that I need to track down still.
Same situation as the last stack leak memory fix, this is fairly tricky
since it is dealing with stack pivoting. Fixes the memory leak around
pthread stack allocations, making memory usage lower for applications
that constantly spin-up and destroy threads (Like Steam).
We need to let glibc allocate a minimum sized stack (128KB and we can't
control it) to work around a race condition with DTV/TLS regions. This
means we need to do a stack pivot once the thread starts executing.
We also need to be careful because the `PThread` object is deleted
inside of the execution thread, which was resulting in a use-after-free
bug.
There are definitely some more memory leaks that I'm still fighting, and I have
noticed in my abusive thread creation program that we might want to
change some jemalloc options to more aggressively cut down on residency.
This is just one out of many.
Previously: Would keep one clone thread's stack active for teardown
delaying.
With aggressive cloning and teardown, this was unsafe.
Only reap the stack when told it is safe to do so.
Which requires moving LinuxEmulation to its own independent folder as
well. Since both IRLoader and FEXLoader rely on it.
No functional change, just moves the the code around.