rlct_clone will acquire a read lock, whereas fork will acquire a write
lock. The write lock is necessary because the fork will clone the file
table, which would result in other threads' fork/clone file descriptors
not being closed. If an address space switch fd never gets closed, fork
child processes and new threads, may never switch address spaces before
they are started, which has resulted in hard-to-debug RIP=0 instr fetch
page faults.

MAP_SHARED means it will continue to be shared after forks, which
obviously isn't the correct behavior. `acid thread` doesn't segfault
infinitely, anymore.

* Correctly align the stack
* Fix argument order of the ldp instructions
* Remove unnecessary ldr x5 instruction