Created initial code for the vanadium_client_sdk; deleted the host crate.

[bigspider]

Initial attempt at separating the host crate's code that should go into the vanadium_client_sdk.

Added a Callbacks trait that allows to inject the V-App-specific logic a V-App client needs to interact with an app execution.

As the current test V-App is extremely simple, the client is currently an executable binary. Real V-Apps will most likely have a proper client library that is not an executable (for example to abstract the V-App's API from its communication protocol).

Closes: #29

[bigspider]

Unfinished things:

• The app should start running immediately, before a message is sent; no reason for the code before the first xrecv to wait.
• The app_hmac is not correctly propagated around (temporarily hardcoded)
• How to get the app's exit status back to the client?
• Didn't test if panics are handled correctly
• How to cleanly stop the V-App if the client exits?

It might be worth trying to restructure the VanadiumClient in order to allow a synchronous version, by 'splitting' the busy_loop into multiple consecutive calls. That is:

• when the V-app is started, run the V-App until the first xrecv;
• then, each send_message processes the xrecv, and runs the V-App until the next xrecv (expecting an xsend in between with the response). Or it stops if exit or a panic is received.

This might simplify building the asynchronous version on top.