Add comprehensive tests to adapter_tests

[lxfind]

This PR adds more tests to adapter_tests.
In doing so, also found a few issues and fixed them:

1. In adapter::publish, we were returning Err for some user errors. We should return ExecutionStatus::Failure instead.
2. In adapter_tests, we were passing MAX_GAS as budget. This is problematic as it equals the the balance of the object, and can fail on the second use (once we add the same check from handle_order to confirm_order).
3. A few nits on error messages.

[huitseeker]

So, if you change the exec_failure! macro to not return something but instead just construct the result, this can become:

let package_id = match generate_package_id(&mut modules, ctx).unwrap_or_else(|e| exec_failure!(..., e))?;

[lxfind]

Not really, note that exec_failure! actually returns Ok instead of Err.

[huitseeker]

Fair enough.

let package_id = generate_package_id(&mut modules, ctx).or_else(|e| exec_failure!(..., e))?;

[lxfind]

That's not going to work either. What I want here is when a user error occurred, the entire function still returns an Ok, but inside that Ok there is an ExecutionStatus::Failure. So any mechanism that chains into ? will not work.
Also exec_failure! needs to be in the outer function scope, not in lambda scope, otherwise it just returns from within the lambda instead of returns the outer function.

[huitseeker]

Yes, understood, finally. You're right, there's nothing relevant to do from within the lambda scope.

[huitseeker]

I'm thinking we may want to provide a convenience method (perhaps an impl TryFrom?) to unpack results wrapping execution statuses.