Finish implementing the wasm3 runtime

[jonas-schievink]

This adds a --wasm3 option to rune run, and duplicates the Wasmer API implementation for wasm3. It also updates to wasm3 0.2, which fixes the soundness bug I encountered last time.

The result:

$ cargo run --bin rune -- run examples/sine/sine.rune --raw .gitignore 
   Compiling hotg-rune-cli v0.6.1-dev (/home/jonas/dev/rune/crates/rune-cli)
    Finished dev [unoptimized + debuginfo] target(s) in 6.83s
     Running `target/debug/rune run examples/sine/sine.rune --raw .gitignore`
[2021-08-23T16:43:56.422Z INFO  hotg_rune_cli::run::command] Running rune: examples/sine/sine.rune
[2021-08-23T16:43:56.689Z INFO  hotg_rune_runtime::common_outputs] Serial: {"type_name":"f32","channel":2,"elements":[27.202106],"dimensions":[1,1]}
$ cargo run --bin rune -- run examples/sine/sine.rune --raw .gitignore --wasm3
    Finished dev [unoptimized + debuginfo] target(s) in 0.09s
     Running `target/debug/rune run examples/sine/sine.rune --raw .gitignore --wasm3`
[2021-08-23T16:43:59.314Z INFO  hotg_rune_cli::run::command] Running rune: examples/sine/sine.rune
[2021-08-23T16:43:59.316Z INFO  hotg_rune_runtime::common_outputs] Serial: {"type_name":"f32","channel":2,"elements":[27.202106],"dimensions":[1,1]}

There are some caveats due to limitations of the wasm3 crate: it doesn't have a very sophisticated interface for building host functions, so I had to manually implement some wrappers that resemble what Wasmer offers here. The crate also doesn't have a way to bubble up runtime errors, which means that an error in a host function results in the WASM binary continuing to run (currently I log the error that occurred and return u32::MAX to the WASM program).

The "all intrinsics are implemented" test from the wasmer implementation has been configured out for now, since wasm3 does not have the required introspection APIs.

[Michael-F-Bryan]

@jonas-schievink, did you want to add the code for propagating errors to this PR before we merge it? Or should I merge it now and we'll add error handling later?

[jonas-schievink]

I'd prefer to land this now so that #271 is unblocked

[Michael-F-Bryan]

I'd prefer to land this now so that #271 is unblocked

Perfect. It looks like you just need to rebase onto master to pick up the latest changes then GitHub should merge this PR automatically.

[jonas-schievink]

@Michael-F-Bryan looks like you need to re-approve the PR for auto-merging to work