Mixed-typed parameters for input words

[mtf90]

While looking through some of the learning related code, I noticed that some of the code expects the type classes of input parameters to be unison. For example:

• The call to teacher.get(t).getFreshValue(old) excepts the old variable to contain DataValues of the same type class (see Theory#getFreshValue). However, the computeOld method adds all constant values to the return value, which may have different type classes (e.g., Integer and String).
• Similarly, the Essentializer gets its fresh values based on all values of the in values which therefore also need to be of the same type.
• The ClasssAnalyzerDataWordSUL#registerFreshValue may return an Integer based value (map.size()) even if the actual retType of the configured MethodConfig is different.

Since all example automata I have found always use the same type class for all parameters (only Integers, only Strings, etc.), I wondered whether this is an intentional constraint or if the above examples are actual oversights/bugs that became apparent with the (new) type system. Should mixed-typed inputs (e.g., a login symbol with a String-based account name but an Integer-based password) be supported by register automata (learning)?