[FIX] Avoid internal corruption of TrieNodes

[brandes-pq]

Unfortunately, I do not have a small query to show the problem, it occured in a very large query that executes a complex test suite.

The symptom was a crash with this message:

Improper use? Potential bug? Your feedback is welcome:
Contact: basex-talk@mailman.uni-konstanz.de
Version: BaseX 12.1
Java: Oracle Corporation, 21.0.2
OS: Windows 11, amd64
Stack Trace:
java.lang.ClassCastException: class org.basex.query.value.map.TrieBranch cannot be cast to class org.basex.query.value.map.TrieLeaf (org.basex.query.value.map.TrieBranch and org.basex.query.value.map.TrieLeaf are in unnamed module of loader 'app')
        at org.basex.query.value.map.XQTrieMap.keys(XQTrieMap.java:44)
        at org.basex.query.func.map.MapMerge.item(MapMerge.java:70)
        at org.basex.query.func.map.MapMerge.item(MapMerge.java:24)
        at org.basex.query.expr.ParseExpr.value(ParseExpr.java:54)
        at org.basex.query.expr.If.value(If.java:159)
        at org.basex.query.expr.gflwor.GFLWOR.value(GFLWOR.java:93)
        at org.basex.query.value.item.FuncItem.invokeInternal(FuncItem.java:131)
        [...]

The general assumption in the code base is that a TrieNode of size 1 is always of class TrieLeaf, see e.g. in XQTrieMap::put() or XQTrieMap::keys().

Removing the second-last item from a TrieBranch has violated this assumption, returning a TreeBranch of size 1. This should be fixed with this patch.