fix(boolean-within): require an interior point in MultiPoint-in-Polygon check

[mahirhir]

booleanWithin(multiPoint, polygon) gives the wrong answer when a MultiPoint has an interior point followed by a boundary point.

isMultiPointInPoly should return true when every point is inside the polygon (boundary allowed) and at least one point is strictly interior. But oneInside was declared and never assigned, so the if (!oneInside) branch ran on every iteration and overwrote isInside with the strict-interior result of the current point. The final return output && isInside then depended only on whether the last point was interior, not whether any point was.

booleanWithin(
  multiPoint([
    [0.5, 0.5], // interior
    [1, 1], // corner / boundary
  ]),
  polygon([
    [
      [0, 0],
      [1, 0],
      [1, 1],
      [0, 1],
      [0, 0],
    ],
  ])
);
// master:   false
// expected: true   (0.5,0.5 is interior, 1,1 is on the boundary, both are inside)

Reversing the points to [[1, 1], [0.5, 0.5]] returns true on master, so the result is also order-dependent.

This is the mirror of booleanContains, whose equivalent helper was fixed for this case in #3024 (booleanContains(a, b) is documented as equivalent to booleanWithin(b, a)). The fix here latches oneInside once a strictly-interior point is found, matching that change.

Added a true fixture covering the interior-then-boundary case (test/true/MultiPoint/Polygon/MultiPointOneOnBoundaryOneInterior.geojson). It fails before the change and passes after; the existing fixtures are unchanged. Not a breaking change.

The open draft #3011 de-duplicates the contains/within helpers and would cover this too. Happy to close in favour of that if you'd rather fold it in, but since the bug is in a released version this seemed worth fixing on its own.

[mfedderly]

Thank you! There's been a bunch of fixes against these de-9im methods. I think you could argue that they're breaking in the sense that we're changing outputs, but they are being corrected to work according to the spec so I think that should be acceptable in a minor release.