bug#24921: 25.1; ERT fails to finish test run if test exceeeds max-lisp-eval-depth

[Gemini Lasswell]

If some code under test by ERT fails by recursing beyond
max-lisp-eval-depth, the resulting error stops the ERT test run without
marking the test as finished or failed.

To reproduce, using emacs -Q or not, enter this code into *scratch*:

(defun go-boom ()
  (go-boom))

(ert-deftest test-go-boom ()
  (should (go-boom)))

And then:
  M-x eval-buffer RET
  M-x ert RET RET

Result: The ERT window appears and shows that it is running the test.
But it never shows that it finishes. Any tests that you have defined
that follow test-go-boom in alphabetical order will not run. The echo
area will show the message "Lisp nesting exceeds ‘max-lisp-eval-depth’"