Update model tester handling around hasChildren()

[The-Compiler]

Relevant upstream changes:

QAbstractItemModelTester: don't rely on hasChildren()
https://codereview.qt-project.org/c/qt/qtbase/+/318767

Dynamic models which use fetchMore to asynchronously fill subdirs
(like KDirModel) return true in hasChildren() for dirs that are expected
to have children (so that the "+" shows in the treeview) but do not
actually have children readily available.
They will be inserted later on once the async listing job is done
(as a result of fetchMore triggering that job).

So QAbstractItemModelTester should use rowCount instead, to find out
if there are children present.

This detected a bug in QConcatenateTablesProxyModel: it returned
a non-zero rowCount for its items, while it's flat.

QAbstractItemModelTester: fix false positive when model has zero columns
https://codereview.qt-project.org/c/qt/qtbase/+/341238
https://bugreports.qt.io/browse/QTBUG-92220

Fix QAbstractItemModelTester false positive
https://codereview.qt-project.org/c/qt/qtbase/+/343539
https://bugreports.qt.io/browse/QTBUG-92886

When rows are removed from a model with no columns,
the test should not report a problem if indexes are invalid

Fix QAbstractItemModelTester false positive
https://codereview.qt-project.org/c/qt/qtbase/+/345195

When inserting rows to a branch with no columns
the tester should not complain about indexes being invalid

---

Still missing one bigger change which I plan to do separately:

Add tests to QAbstractItemModelTester checking only one change in flight (I15fe1ad1) · Gerrit Code Review

[nicoddemus]

Took a glance at the patches you posted (thanks for the detailed description, as always) and the code changes, everything LGTM. 👍

[The-Compiler]

Build failures are unrelated (see #419), merging this so I can submit part 2 😉