Refactor block validation multiple pipe inputs

[georg-schwarz]

• Changed logic to iterate over pipeline structure.
• Adapted some tests to be more precise

Validation output should have stayed roughly the same (except some minor edge cases)

[georg-schwarz]

A discussion emerged from this regarding #485 between @joluj and me. Probably of interest to @rhazn
For the sake of completeness, here a condensed version. The discussion needs to be continued somewhere else I think.

We currently don't model instances. This works until now as we prohibit scenarios where a block is reused by different inputting blocks (this validation was refactored in this PR). The question is if this is really desirable. If we allow it, we might run into ambiguous behavior.

Our takeaway was that we need to think ahead about how we want to model blocks with different ports (e.g., for merging different tables). This seems to be the stage where we need to model instances. We did a whiteboard session on this, and it feelt really difficult to do without an instantiation logic where you can indicate whether a block is the same instance or if there are different instances of a block. This is probably the origin for this feedback.