Improve opt passes again

[MaxGraey]

This also speed-up passing tests

[dcodeIO]

Would be great to have even more comments around there as it's hard to tell what's going on, and why, and in the exact order. Worrying that we'll otherwise accumulate pass-bloat, forgetting why specific passes are at the location they are, becoming impossible to reason about questions like "do we actually need to run this pass here?".

// now that stuff-X has been reduced by pass-X we can optimize the resulting stuff-Y with pass-Y, doing explain-Y
add("Y");

Questions to ask to make a good comment:

• What does this pass do?
• Why at this location?
  • Which passes does this depend on?
  • Are there passes that should not be run before?
• Why limited to this optimize or shrink level? (if applicable)
• Why do we run this pass another time here? (if applicable)

[MaxGraey]

Sure. I will try to describe all my steps, but it not always logical, sometimes just empirical find outs.

Btw I'm wondering is SSA used for globals? I guess it could optimize much more. See this for example:
https://github.com/AssemblyScript/assemblyscript/blob/6326ddf95408ac62775b24f7328c51235263d758/tests/compiler/logical.optimized.wat

[dcodeIO]

Afaik (global ..) optimizations are relatively basic, limited to detecting single uses that are never reassigned and such. Especially C/C++ don't emit a lot of globals, so I guess it hasn't been a high priority.

[dcodeIO]

In 005dfbc I rebuilt the allocator tests and noticed that though the untouched fixtures became a bit smaller, the optimized ones became larger. May this have something to do with pass order or is this unrelated?

[MaxGraey]

tests/allocators don't update pretty often. See for example latest changes:

  (type $FUNCSIG$iii (func (param i32 i32) (result i32)))
  (type $FUNCSIG$ii (func (param i32) (result i32)))

to

 (type $i32_=>_none (func (param i32)))
 (type $none_=>_none (func))

and that old stub version didn't has maybeGrowMemory for example and other extra code which happened during long period

[MaxGraey]

@dcodeIO btw I updated that tests in this PR as well and it significant reduce sizes:
https://github.com/AssemblyScript/assemblyscript/pull/1045/files#diff-c18c517ac7bd11043db66de8d39266de