BASELINE 1.0.12.7 (unaltered released ZIP)

test_regression: RESULT cases=101 passed=101 failed=0 assertions=1341
test_helper_menu: RESULT cases=65 passed=65 failed=0 assertions=2783
test_crop_support: RESULT cases=80 passed=80 failed=0 assertions=3100

All 246 previous cases pass in the unchanged baseline. They do not establish
correctness for a native unset windrow flag. The earlier crop fixture modeled
no-windrow crops as hasWindrow=false and native reset as ==false. The actual
user log contains an identified CANOLA descriptor with hasWindrow=nil.

Changes to tests in 1.0.12.8:
- Native reset model now checks not fruit.hasWindrow, matching published GIANTS code.
- One old expectation "missing hasWindrow stays unknown" on a fully identified
  fruit is replaced with explicit no-windrow classification for that case.
- 56 additional cases cover native nil flags, malformed/unknown descriptors,
  source order, state restoration, genuine manual override, and diagnostics.
- Existing regression and helper-menu test files are byte-identical to 1.0.12.7.
