From PR #1824

* This PR is for PMD 7
* It's extracted from #1759

Changes:
* adds the new `assertTextRangeIsOk`, that is executed on every node in
* the kotlin-based tests
* this ensures, that the position of the nodes in the AST is the same as
* they appear in the source code
* For the new switch labeled rules, we have the following grammar

```
void SwitchStatement():
{}
{
  "switch" "(" Expression() ")" SwitchBlock()
}

void SwitchBlock() #void :
{}
{
    "{"
        (
            SwitchLabel()
            (
                "->" SwitchLabeledRulePart() (SwitchLabeledRule())*
            |
                ":" (LOOKAHEAD(2) SwitchLabel() ":")*
(BlockStatement())* (SwitchLabeledStatementGroup())*
            )
        )?
    "}"
}

void SwitchLabeledRulePart() #void:
{checkForSwitchRules();}
{
    (
        ( Expression() ";" ) #SwitchLabeledExpression(2)
      |
        ( Block() ) #SwitchLabeledBlock(2)
      |
        ( ThrowStatement() ) #SwitchLabeledThrowStatement(2)
    )
}
```

`#SwitchLabeledBlock(2)` takes the last two nodes (SwitchLabel + Block)
and adds them as SwitchLabeledBlock to SwitchStatement. JavaCC sets the
first token of SwitchLabeledBlock to be the Block node, but it should be
the SwitchLabel node.

This is fixed in the `jjtClose` methods in the three related
SwitchLabeled* nodes.

On the expression grammar PR, there is a better solution (you can mark
the nodes via the interface `LeftRecursiveNode`).

* So, this actually fixes a bug, that is present in PMD 6.