Uniformize associated data and entry point inputs

[fjarri]

After writing malicious behavior tests for a real-life protocol (CGGMP), it is evident that ProtocolError::AssociatedData is pretty much all the shared data (is it the same as all the public data?) that is passed to an entry point constructor. Should we make it official somehow?

[fjarri]

After #117, we will have SharedData in the Protocol. It is still an open question whether it makes sense to use it in EntryPoint::make_round() (adding a SecretData associated type).