optimize 32-bit x86 system calls

[andrewboie]

Is your enhancement proposal related to a problem? Please describe.
System calls on 32-bit x86 use the old-style int 0x80 mechanism to trap to supervisor mode (Ring 0). There are more efficient ways to do this.

Describe the solution you'd like
Use SYSENTER/SYSEXIT instead to perform syscalls. This has much less overhead but requires manual stack switching. Our 64-bit port already uses the very similar SYSCALL/SYSRET.
https://c9x.me/x86/html/file_module_x86_id_313.html
https://c9x.me/x86/html/file_module_x86_id_314.html
https://wiki.osdev.org/SYSENTER

Describe alternatives you've considered
None.

[andrewboie]

Top two commits of this old branch is where I started working on this, but I set it aside. https://github.com/andrewboie/zephyr/tree/sysenterexit