Implement looping about lineiform

[chc4]

Once we have this I'd like to do a bunch of cool optimizations like loop invariant code motion and intelligent loop unrolling. They're probably really hard to implement without a CFG or DFG, though, which is a Future Chc4 problem.

[chc4]

Instead of doing 1 or 2, we do technically have a third option: do an initial "slow" instrumented execution of the function while collecting trace information, and do lazy basic block version lite where we only compile the path we have branch information from (with adding new branches if the JIT tries to run them).

This is how redmagic works, along with any tracing JIT (including Graal and RPython and Higgs and LuaJIT and...). I don't really want to do it since Lineiform is mostly geared towards a method based JIT instead of a tracing JIT for now, collecting trace information is kinda hard (you either need to single step emulation of instructions, using something like https://github.com/lifting-bits/microx or Unicorn, or Intel PT which is processor specific), and the lazy part of LBBV means you have to call back into and continue JITing from potentially any branch.