refactor(*): refactor wasm bytes preparation logic to the core crate

[Mossaka]

This PR refactors the common Source logic from runtime shim to the core crate. To do so, the core Engine trait API has a breaking change:

1. run_wasi() now takes a wasm_bytes: &[u8] parameter as the wasm module / component loaded from the file system or OCI layers to execute.

cc @jsturtevant

[jprendes]

I think I would prefer if Source provided a method called to_bytes, that would encapsulate this logic, rather than changing the run_wasi signature.

[jsturtevant]

The background is that we had something similiar to to_bytes but this had a clone #401 (comment). Maybe I've got it wrong in that thread about trying to avoid a clone on that call?

I am also not a fan of these changes to the run_wasi trait

[jprendes]

The clone could be avoided returning a Cow<[u8]> IIUC.

[jprendes]

Alternatively, in the case of File the bytes could be cached to avoid reading the file multiple times (something like a OnceCell or OnceLock), and then the &[u8] would refer to the bytes owned by the cache.

[Mossaka]

@jprendes thanks for the suggestion of Cow! I re-implement the refactor in #454

[Mossaka]

Close in favor of #454