the idea is to make libtf completely multi-threaded. meaning each
fiber can be running concurrently in separate thread. quite a bit
of framework is added for this and some atomic helpers are already
introduced. however, io polling is busy polling now (will be soon
in own thread) and timeouts are still more or less broken. oh, and
the multithreading core is not there yet. basically we are currently
mostly broken ;)

use it for heaps and fiber stacks.

so we can in future have more polling frameworks than epoll.

debugging looks now better. there was also some valgrind issues
that needed attention.

it's less confusing this way.

by letting gcc figure out better how to represent stack pointers.
also uses labels to make the code more readable.

it's faster and has smaller context in the beginning of the
fiber. it's also required since setjmp() seems to use mangled
pointers in glibc so we could not schedule fibers in non-creation
threads. additionally the setjmp() setup code has a race condition.

instead of having per-function argument, use a push/pop mechanism:
- multiple timers inside fiber use only one heap entry
- easy to chain multiple possibly blocking operations inside one
  timeout block