Docs: further documentation of static features

[anshumanmohan]

I have made progress on documenting the static qualifier, but have left out some advanced magic since it may belong elsewhere.

@sampsyo has summed up the TODO nicely, so I'll let him take it away:

I have put nothing in there about refinement; it's perhaps "just a cute detail" as far as the user is concerned. Thoughts?

I think it would be great to add something about this someday, but in a separate PR! Namely, a thing that users might want to know is the following paradigm for how they might use static:

1. Emit fully dynamic code everywhere.
2. Observe places where performance is suffering because static inference was incomplete.
3. Sprinkle static qualifiers in there.

The relevant idea offered by refinement is that step 3 is guaranteed not to break the program.

By this logic, I'd say that inference should be presented here, but I'm not sure about compaction. Thoughts?

Yep!! In particular, maybe we can talk about inference/promotion when we add stuff about refinement above. It's good for people to know this pass exists; the details of how it works or how capable it is probably are not worth writing down. Compaction seems like "just an optimization" like any other, i.e., it can be documented like any other optimization pass.

Originally posted by @sampsyo in #1872 (review)