fix slow compilation by using dynasm VecAssembler

[ailisp]

Compile was slow, especially for many function contracts because it use dynasm Assembler, which doing a few system call per wasm function. These system calls are about memory management (mmap, munmap, mprotect), but these are not used by wasmer at all (great finding by @matklad !). Drop them make compilation 10x faster:

Before ( 093eb59 ):

cargo run --release -p wasmer-workspace --example compiler-singlepass --features singlepass
    Finished release [optimized] target(s) in 0.12s
     Running `target/release/examples/compiler-singlepass`
code.len() = 1456
Compiling module...
  2.35s            Module::new (compile)

After:

code.len() = 1456
Compiling module...
  235.85ms         Module::new (compile)

[syrusakbary]

This is awesome. We knew using a VecAssembler could be more optimal, but we didn't measure by how much... so great work!
Do you mind to do a PR also upstream @ailisp ? Thanks!

[ailisp]

@matklad ptal :)

[matklad]

LGTM, and thanks for upstreaming this! Let's hold off pulling this in until the next week -- we are going to branch the release soon, merging after release should give this change more time to bake in, and maybe even update the corresponding costs.

[nagisa]

I went ahead and manually cherry-picked in the change from upstream at 7945c3e, since its been approved and landed there.