feat(compiler)!: Block off 1KiB of heap space as generic swap

[peblair]

Currently, there is no way to allocate heap space in WASI polyfills (since they cannot import the dataStructures.gr module). To support cases in which some minimal heap use is desired (e.g. the NEAR SDK), this PR blocks off 1KiB of memory (0x400-0x800) which WASI polyfills can freely write to.

(Technically breaking, since the location of the runtime memory is moved)

[phated]

This is a workaround for #841, right? Did @ospencer have any ideas on how to allow memory allocation in the polyfills to avoid this breaking change?

[peblair]

It does appear to be the same issue, but I had come up with this independently. This PR is a workaround for that issue. If @ospencer has a better idea, we can do that, but I need some sort of solution for the NEAR SDK.

[ospencer]

Not sure how I feel about this yet, but in terms of something that'd work immediately, you can use a string or an array to block off some swap space.

[ospencer]

@peblair I was chatting with Blaine about primitives and I had an idea about this:

Right now all of our primitives are functions, but I think we could have some values as well! I was thinking through that wasm4 issue and philip's PR for reserving a chunk of heap space for something that needs it—I think it's a good use case for a @heap.base primitive that you can populate with like a --memory-base to scoot up where we start the heap. Then you can reserve whatever amount of space you need for whatever your use case is, would set us up to support something like AssemblyScript's memory.data (which would be so nice) etc.

He liked that idea. If you think that sounds reasonable too, I can make an issue for it!