Implement performance optimizer

[64kramsystem]

Super-simple™ implementation of the performance optimizer.

This is a per-instruction optimizer, which is in turn a simplified optimizer in itself. Future extension to multiple-instruction optimization will surely require a more extensive change.

There is a metric ton of design considerations. This PR can be considered a draft, or not :) The most important considerations are:

• the Fisk::Instruction fields have been exposed; this is necessary, as the optimizer needs to know the instruction, and will need even more, in case it will work on multiple instructions; the alternative OO-design of adding an optimizer on the specific Fisk::Instructions::Instruction is bound to break down
• the Optimizer could now absorb the performance check (performance checking is now located separately from performance optimization, which is not Vewy Nice); I think it requires a bit of design agreement (I have ideas about what to do), which can be done either in this PR or in a separate one.

Relates to tenderlove/tenderjit#49.

[64kramsystem]

Ah, I've added a commit that cleans up a warning, avoiding a separate PR. If you prefer strict PRs, LMK, I have no problem with that approach 🙂.

[64kramsystem]

@tenderlove Do I need to do anything on this PR? Ultimately, more significant changes (e.g. merging checking and optimization logic) can be done separately (or not done at all!).