Refactor fuzzing test

[Leeeon233]

This is a previous proposal, but the code has never been implemented.

Now use Map as an example.
There's a lot of redundant code and some inconsistencies in the current fuzzing tests, so I've introduced an additional abstract. The purpose of this is to:

• Minimize boilerplate code.
• Facilitate the extension of new semantic types, such as movable list, table, etc.
• Simplify the extension and maintenance of common actions, like sync and checkout.

What are your suggestions?

TODO

• new fuzz crate
• add List Map Text and Tree in the new fuzz
• add checking that deep value is equal

More ergonomic action and error output for debugging

[zxch3n]

Can we move the fuzz code to a new crate like fuzz? I guess they don't need to rely on the internal stuff now, correct? It was inside loro-internal because we needed it to test some internal implementation back then. Then, we may remove the test_utils feature in the future. And we can now rely on loro instead of loro-internal.

[Leeeon233]

Can we move the fuzz code to a new crate like fuzz? I guess they don't need to rely on the internal stuff now, correct?

I think so.