Rationalise compiler ops

[kitsonk]

This PR removes the code cache OP from the runtime and rationalises the response from the compiler. Currently, the compiler performs a cache op and then returns the full module (including the source code) back to Rust. Rust already knows the source code and knows it needs to try to cache the result after a compile, so it handles all of this, removing the need for the compiler to perform these actions.

Because I am still a Rust infant, I am sure I made mistakes in the Rust code. I still need to restore the tests and write new ones.

I made the decision to rename CodeFetchOutput to ModuleMetaData because that seem better to represent what is happening, as the object doesn't always come from a code fetch and is incrementally enhanced after the compile.

[ry]

I'm slightly worried we may want CodeCache in the compiler isolate again at some point in the future - but I think it's good practice to remove dead code.

waiting for signature change on compile_sync

[kitsonk]

Ok, the PR now does the following:

• Moves the caching of the compiled output to Rust from compiler runtime. Removes the code_cache op. The compiler was sending back all the data anyways in compile_async, so it was effectively doing it twice.
• Converts the code_fetch op to fetch_module_meta_data and moves the source code from a string to an 8-Bit array.
• Generally migrates source code, output code and source maps from String to Vec<u8> within Rust.
• Tries to make the term ModuleMetaData consistent for objects that represent a bundle of a modules descriptor/source code/compiled code/source maps, be it in Rust or the compiler runtime.

[ry]

LGTM - thanks Kitson !