[Feature] Implement Buffer.alloc and Buffer.allocUnsafe, add tests

[jtenner]

Notes:

• Using toThrow() is nice here because we can assert the module will panic (and be caught inside as-pect)
• Downside: throw does not exist in assemblyscript and it checks for unreachable() instead (we can't determine why the unreachable() occurred)

afterAll(() => {
   expect<i32>(RTrace.count()).toBe(value);
});

• Great reuse of E_INVALIDLENGTH here

Thoughts?

[jtenner]

I like this solution better. Count the references after each test. For some reason #allocUnsafe has 12 instead of 14, but it's nice to set a fixture like this.

[dcodeIO]

Not sure what to do with the throw asserts, since everything that can be done at this point is suboptimal. Suggesting to ignore these things for now, because given the complexity this adds, asserting that an u32 comparison for example works doesn't exactly justify the effort.

A comment to improve this eventually, once exception handling becomes a thing, is great, of course.

[jtenner]

Not sure what to do with the throw asserts, since everything that can be done at this point is suboptimal.

I don't know exactly what you mean by "suboptimal."

If you held my feet to a fire, I wouldn't mind leaving out the throw asserts. It does seem to be good enough for the std lib. There is also another part of me remembers the many bug fixes I had to make in jest-canvas-mock because I didn't follow the specification correctly and didn't throw errors where they needed to be thrown.

As for the RTrace stuff, we can definitely assume that AssemblyScript does the job correctly and remove those too.

[dcodeIO]

With suboptimal I meant that all things throw in AS don't work properly anyway yet, thus requiring workarounds, respectively that baking reference counts into the tests is susceptible to breakage if something on either the compiler or optimizer side changes. Seems hard to maintain.

[jtenner]

Yeah. The toThrow() assertion merely looks for unreachable(). I know it's suboptimal, but as-pect will support try/catch the moment AssemblyScript does, and I see no reason why we should avoid using it, despite being not optimal.

We should definitely remove the RTrace stuff.

[jtenner]

Okay. I think the functions are ready for review. Anything that needs to be changed?