protect from cyclical sh:path

[VladimirAlexiev]

What will happen if an engine encounters a malformed sh:path that has a cyclical link to itself?
A naive engine may just loop forever, which constitutes a simple Denial of Service attack vector.

• https://www.w3.org/TR/shacl/#property-paths says
  "A node p is not a well-formed SHACL property path if p is a blank node and any path mappings of p directly or transitively reference p."
  • I suggest to add:
    "Implementors are encouraged to check for such cyclical property paths and raise a failure, because such paths constitute a potential Denial of Service attack vector"
• (Note: https://www.w3.org/TR/shacl/#shapes-recursion says
  "The validation with recursive shapes is not defined in SHACL and is left to SHACL processor implementations."
  Although somewhat related, this is a completely different issue.)
• SHACL-SHACL doesn't (and cannot) check for cyclical paths, see shsh:PathShape (although it uses a named blank node _:PathPath).

We should also add a test for it.
Out of all path tests https://github.com/w3c/data-shapes/tree/gh-pages/data-shapes-test-suite/tests/core/path
I can find only https://github.com/w3c/data-shapes/blob/gh-pages/data-shapes-test-suite/tests/core/path/path-unused-001-shapes.ttl#L12 that has a cyclical path _:p2:

_:p1 sh:inversePath _:p2 .

_:p2 sh:zeroOrMorePath ( _:p2 ) .

_:p2 sh:zeroOrOnePath [ rdf:rest rdf:nil ] .

_:p3 sh:alternativePath ( ex:p ) ;
  rdfs:comment "invalid path" .

But it's not used in any node shape, so it does no harm.

We can change the node shape slightly to use the cyclical path:

ex:s1 a sh:NodeShape ;
 sh:targetNode ex:j, ex:i ;
 sh:class ex:C;
 sh:property [sh:path _:p2] .

Of course, it's better to write it as a separate test path-cyclical-001