fix(report): the console summary counted the wrong skips

render() computes the catalog's own skip count, then the scope paragraph
rebound the same name to the run-wide census:

    skipped = sum(... for t in s[5] if _lookup(...) == 'skip')   # catalog
    ...
    skipped = JUNIT_CENSUS['skipped']                            # whole run

so the line CI operators actually read --

    N sections, 374 tests (323 passed, 0 failed, 3 skipped, 48 pending)

printed the whole run's skips inside a breakdown of the catalog. The four
numbers did not add up to the total, and the error ran in the alarming
direction: it inflates skips, which reads as "lots of this was not exercised"
against a catalog where only three entries were actually gated.

Use the census value inline where the paragraph needs it and leave `skipped`
meaning one thing.

Added the assertion that would have caught it, since the breakdown is only
ever right when it reconciles:

    assert passed + failed + skipped + missing == total

Now: 374 tests (323 passed, 0 failed, 3 skipped, 48 pending) -> 374.

Author: BitHighlander

scripts/generate-test-report.py

@@ -2845,13 +2845,12 @@ def _section_state(s):
     ran = JUNIT_CENSUS['ran']
     if ran:
         pb.gap(3)
-        skipped = JUNIT_CENSUS['skipped']
         for line in _w('Scope: this report is a curated catalog of %d tests. The CI run collected %d '
                        '(%d of them native firmware unit tests); %d SKIPPED and did not execute, '
                        'usually because the emulator predates the firmware the test targets -- a skip '
                        'is not evidence the feature works. Absence from this report is NOT '
                        'evidence that a feature is untested -- check the JUnit artifacts.'
-                       % (total, ran, JUNIT_CENSUS['native'], skipped), 100):
+                       % (total, ran, JUNIT_CENSUS['native'], JUNIT_CENSUS['skipped']), 100):
             pb.text(8, line, color=GRAY)
     pb.gap(6)
     pb.text(12, 'Sections', bold=True)
@@ -2994,6 +2993,9 @@ def _section_state(s):
 
     pb.finish()
     pdf.write(output_path)
+    assert passed + failed + skipped + missing == total, (
+        'catalog counts do not reconcile: %d+%d+%d+%d != %d'
+        % (passed, failed, skipped, missing, total))
     print(f'{output_path}: fw={fw_version}, {len(active)} sections, {total} tests '
           f'({passed} passed, {failed} failed, {skipped} skipped, {missing} pending)')