Fix (or replace) `doc/ref/testconsistency.g`

[ThomasBreuer]

When doc/ref/testconsistency.g is read by GAP, a lot of output appears that claims to exhibit errors in the GAPDoc syntax of the GAP manual.
Some of these claims are wrong, for example "Transitivity uses Oper instead of Attr" -- Transitivity is both an attribute and a non-unary operation, and one wants to be able to refer to the non-unary situation.
Luckily, most of the output is irrelevant for the status value that is used in a CI test.

It would be already an improvement if the wrong output lines would be omitted.

A better solution would be if the idea of "more strict checks" would become part of GAPDoc.
The decision about "officially" correct GAPDoc syntax should be left to GAPDoc, perhaps via an optional argument that causes a more strict check of references etc.
Such an official "strict mode" could be used also by package manuals based on GAPDoc, if one really wants this.
Perhaps @frankluebeck has an opinion about this.

[ThomasBreuer]

Update after a discussion with @frankluebeck:
The code in doc/ref/testconsistency.g tries to match elements from a GAPDoc document with objects from the current GAP session, this is beyond the scope of GAPDoc.
It is not clear which consistency rules in this sense are useful.
Maybe it is sufficient for the moment to remove the wrong parts and to suppress the output that is irrelevant for the test result.

[fingolfin]

I've never quite understood what we (resp. GAPDoc) even gain from distinguishing Func/Prop/Oper/Attr/... that makes the extra complication useful? OK, if here is both an attribute and an operation with the same name, then one can use this to distinguish them. But that could also be done with a label?

Ah well, I guess it is deeply baked in by now.

Regarding doc/ref/testconsistency.g: If it helps us find broken refs, great. Otherwise: it is not even clear to me what breaks if one gets this wrong?

[ThomasBreuer]

Given a GAP operation, there can be many ManSections that provide information for it, and if the operation is also an attribute or property then some ManSections are about the attribute or property, and it makes sense to state this in the documentation because then one sees whether the values will be cached and can be used in the method selection.

For cross-references in GAPDoc documentation, one can use labels to address individual ManSections.
Thus the distinction between Func, Oper, Attr, Prop in Ref elements is not relevant for the structure and correctness of the document. In fact, we could always write <Ref Func="..."/> in these cases.

When GAPDoc processes a document and does not find broken cross-references then also doc/ref/testconsistency.g does not find broken cross-references, it just complains about different names used in a Ref and in the element to which it points. If we agree to remove doc/ref/testconsistency.g then I do not object. The file had been written by @olexandr-konovalov, perhaps he wants to comment.