[General] refactor the type-mapping logic to clearly separate build-in types from API-defined types

[ePaul]

Swagger Codegen has some logic which selects type names for generated models and for properties.
This handles both mapping to build-in types of the programming language (e.g. for Java int/double/String/BigDecimal/UUID/...) and name conversion (e.g. case change, escaping for reserved words, model name prefix/suffix) for types defined in the API definition file (mainly object and enum types).

Currently the code is a bit convoluted, which leads to bugs like #3678 (where the suffix/prefix is applied to a build-in type).

Solution proposal

This logic needs to be refactored to separate both functions clearly. Any name modifications should only be applied to for API defined types, not to build-in types of the programming language.
(I guess handling of maps/lists would also be part of this.)

[ePaul]

A proper fix for this would also make sure that in generated documentation (in markdown) there are links only to generated classes, not to ones included in the environment (and not documented).

Noted in #5118 (comment).