fixing notebook test errors when no tests collected (quantumlib#4196)

Fix broken notebook tests in master.

It seems that when `pattern` does not match any tests in `pytest -k pattern`, pytest returns an exit code 5. This breaks the newly sharded notebook tests, which are leveraging the `-k partition-n` pattern for changed notebooks. 

This fix adds the fix as per the recommendations of pytest-dev/pytest#2393.

check/pytest

@@ -41,9 +41,17 @@ fi
 
 if [ -z "${ACTUALLY_QUIET}" ]; then
     pytest "${PYTEST_ARGS[@]}"
+    RESULT=$?
 else
     # Filter out lines like "...F....x...      [ 42%]", with coloring.
     pytest -q --color=yes "${PYTEST_ARGS[@]}" |
         perl -nle'print if not m{^(.\[0m)?[\.FEsx]+(.\[36m)?\s+\[\s*\d+%\](.\[0m)?$}'
-    exit "${PIPESTATUS[0]}"
+    RESULT=${PIPESTATUS[0]}
 fi
+
+if [ "$RESULT" = 5 ]; then
+  echo "[exit 5] No tests collected, but ignored."
+  exit 0
+fi
+
+exit "$RESULT"