This could 'easily' reproduced by opening a large git-based project in
kdevelop with one of the AbstractFileManager-inheriting project
manager. Then switching between branches resulted in a crash sooner or later.

The main culprit was that FileManagerListJob::slotResult was sometimes
called after a the job was killed. KJob::kill apparently does *not*
kill the job directly, hence we need some additional safeguard here.

Furthermore this patch adds some more safety measures against invalid
item pointers in the FileManagerListJob's m_listQueue.

Please test!

BUG: 260741