bug#51941: Detect duplication of ERT tests

[Glenn Morris]

Mattias Engdegård wrote:

> It would certainly be good to detect such mistakes, but I'm not sure
> how to go about finding them. I wrote a quick-and-dirty scanner that
> found about 400 uncalled functions in the test tree of which 1 or 2,
> maybe, should have been `ert-deftest` (will fix those right away).
>
> Clearly something with much better signal/noise ratio is called for.
> One would be to only consider functions using `should` etc.

Seems to me that making "should" only be defined within tests would have
flagged these (https://debbugs.gnu.org/28385).