Fix a minor todo-mode regression

Fix a minor todo-mode regression

* lisp/calendar/todo-mode.el (todo-get-overlay): Wrap in
save-excursion.  This fixes a regression introduced by the fix
for bug#27609, whereby trying to raise the priority of the
first item or lower the priority of the last item, which
should be noops, moves point to the item's start.  Clarify
comment.

* test/lisp/calendar/todo-mode-tests.el
(todo-test-raise-lower-priority): Add test cases for trying to
raise first item and lower last item.
(with-todo-test): Clear abbreviated-home-dir, since we change HOME.
(todo-test-toggle-item-header02): Remove ":expected-result
:failed" and tests of point after todo-next-item, since the
effect when using Todo mode is not reproducible in the test
environment.  Add commentary about this.