[Documentation] Error handling new way
The documentation describe that in an ideal world the message file should be complete. However for why3 where the language is not fixed, it is hard to keep that property ( why3/why3@ce21b6c1 ). However the new error handling way is a lot simpler than the old.
Could the documentation state what is lost when completeness is not checked? I believe we just lost that we could believe that we are handling a specific parsing error, but in fact it could append in other states.