Implement fuzz testing for Wasm execution

[Robbepop]

We currently have a very naive fuzz testing infrastructure for wasmi that works and already found bugs.
However, this fuzz testing infrastructure currently only tests Wasm parsing, validation and Wasm to wasmi bytecode translation but does not cover the important Wasm execution phase.

For this we require another Wasm engine that runs on the same Wasm inputs as wasmi. After each run we compare the state of the virtual machines (globals, linear memories, execution results, etc.) and see if they match. Divergent states indicate differences in execution behavior and therefore a bug on one side or the other.

The other engine should have a very strong focus on correctness and Wasm conformance. Nothing could beat the official Wasm spec interpreter in this regard.
Rust bindings for the official Wasm spec interpreter already exist but there currently has not been a public release on crates.io.

Once the official Wasm spec interpreter has been released on crates.io we can make use of it in a similar manner as done in Wasmtime.

[Robbepop]

Ideally we want two different fuzzing strategies to fuzz wasmi execution:

1. Fuzzing execution for the register-machine wasmi engine.
   • This one just guards against unwanted panics at execution time.
2. Differential fuzzing against the stack-machine wasmi engine.
   • This one asserts that executions between the stack-machine and register-machine wasmi engines are the same.

[Robbepop]

This PR (#832) implements execution fuzzing for wasmi.