Figure out situation regarding resetting the runtime Wasm VM

[tomaka]

At the moment, virtual machines are re-used when executing multiple runtime functions in a row (even across blocks).
However, the memory of the virtual machines isn't zeroed, for performance reasons.
The allocator, however, is re-recreated between each call. This might be problematic because the allocator works by writing data in the memory. The allocator might assume that memory is in a certain state when it's not.

The whole situation should be figured out. Is it even correct to not reset the memory?

cc #132

[tomaka]

Another important thing to note is that the Wasm instance also contains a stack and globals, which need to be reset as well.
This means that we should recreate the Instance from scratch every time no matter what.

[tomaka]

Some notes from lil pep:

if you use wasmtime, use the pooling allocator. If not, and you are on Linux (not other nixes), use madvise(DONT_NEED), if you are on macOS help yourself or maybe Alex T. might help you. If you are in browser I guess the best way is to rely on the browser engine if possible. If not, I wonder maybe web supports quick tricks on buffers.