The `call_guest_function_by_name` method changed behavior

[jprendes]

How do we ensure when someone upgrades to this new api that they don't assume that memory was already restored when it actually wasn't? Is that a concern?

Originally posted by @jsturtevant in #697 (comment)

In #697 we changed the behaviour of call_guest_function_by_name, which can catch users by surprise. The linked issue proposes different alternatives to address this issue.

Some of the options proposed:

• Ensure we communicate this change clearly, using rustdocs and release notes
• Use the new behaviour in a call method, and replicate the old behaviour with the new API in the call_guest_function_by_name
• Same as before, but with a deprecation warning on call_guest_function_by_name
• Same as before, but with an unconditional error instead of a deprecation warning

[jprendes]

We should address this before the next release

[simongdavies]

I think we should investigate the impact of this before we make a decision, but, my preference would be that we revert the behaviour of the API so that it does not change and introduce a new API that has the new behaviour.

hyperlight-js uses the old behaviour (as I think hyperlight-wasm does at least for modules). I would imagine that @qbradley was relying on the old behaviour although I could be wrong)

[simongdavies]

When we make these changes we should also reflect the changed functionality in the micro-benchmarks

[syntactically]

We agreed that we would revert the behaviour of the old API, mark it deprecated, and introduce the new API under a different name.

[jsturtevant]

#761