getFaceByPoint test relies on invalid topology

[strk]

Trying a faster Point-In-Polygon test I found that topology/test/regress/getfacebypoint.sql is expecting a debating result for the test 't5'.

The argument point falls within an "inner" face but the "inner" face isn't correctly registered but rather marked as being part of the universe face. This is an invalid topology, constructed via AddEdge/AddFace calls within the test itself.

I suggest the topology is corrected and both a test for point-in-hole and point-in-universe replace the existing 't5' test.

Andrea: adding you in Cc as I believe you were the author of that test

[strk]

In 15129:

Fix test for getFaceByPoint to not rely on invalid topology

Closes #3632