Byte code

[udoprog]

Add preliminary support for building and loading byte-code into the Vm from raw bytes, instead of having instructions always being an array of Vec<Inst>.

This has the potential of greatly reducing the size of Rune programs, but it comes with a performance penalty because instructions have to be decoded as they're being executed.

By default, we're still using the old array-based instructions format.

See the difference in in-memory size for the following program:

use std::iter::range;

pub fn main() {
    let it = range(0, 1000);
    let tail = 77;

    'label:
    while true {
        let value = 10;

        for n in it {
            loop {
                let value2 = 20;
                break 'label;
            }

            tail = tail + 1;
        }

        tail = tail + 1;
    }

    tail
}

Byte-code version: 196 bytes.
Instruction array: 2560 bytes.