Further refine the Complexity metric

[ResistantBear]

There are still points of disussion left, including:

• [ ] Whether to include throw-statements. Currently, we do no longer include them, like Sonar does for Java. Sonar does include them for JS/TS, however. So we cannot use Sonar as orientation here. Not counting for them is closer to the original MCC definition, as they do not branch the code, but unconditionally jump to somewhere, just like a goto.
• [ ] Whether to include return statements and goto statements. We currently do not do so and Sonar does seem to only include GOTO statements in a language where they can be conditional. Not counting for them is closer to the original MCC definition, as they do not branch the code, but unconditionally jump to somewhere.
• [ ] Whether to include catch-blocks. As they can cause a branch of the code based upon the type of the thrown error, they probably can be counted when following the original MCC definition, but that whole throw/catch-thing is rather tricky when it comes to the control flow. Currently, we count them.
• [ ] Whether or not to count the conditional C(++) preprocessor directives like #if, #ifdef or #ifndef, which are covered by the two syntax node types "preproc_if" and "preproc_ifdef". As they are only evaluated at compile time, we might continue to not count them.
• [x] Whether to include method references (like in Java), since they counts as method. In some cases, they are the same as lambda expressions => No, see comment below
• [ ] Whether to count OR in switch-cases, e.g. case "a" or "b", which is supported by some languages, or not. As far as I know, we do count that in languages where this OR is handled by one of the Syntax Nodes which are also used for OR on other places like C#, but not for Rust, Python or Ruby. If we try to be consistent to what is done for regular OR and AND expressions, we should also count that. However, either way, we probably require more special handling again to make this consistent across programming languages.
• [ ] Whether to count when case guards in switch labels (C#)
• [ ] Whether to count build-in functions used for (more) functional programming, e.g. .forEach or .catch in JS/TS or forEach and map from Java Streams. Note that a user could also build his own foreach function using recursion, etc., and we are unable to check for such a thing.

[ce-bo]

I put it to state "new issue" since we need to prepare/refine the issue.

[ResistantBear]

Method references seem to be the Java-concept for passing a function by value. As that does not involve implementing a new function (at least not from the view of the programmer), we should not count them as new function definition.

Originally posted by @ResistantBear in https://github.com/MaibornWolff/metric-gardener/issues/94#issuecomment-1934219476