Add grid test 769: inner ring touching outer at two points, splitting it into two components

[brawer]

grid has fixtures for an inner ring touching the outer at a single point (7/754-7/758) and touching along a shared line (7/756, 7/757), and for a single self-touching way resolving into two components (7/765, 7/766). It doesn't seem to have a case for two separate simple rings — one outer, one inner — where the inner ring touches the outer at exactly two points, so removing it splits the outer's boundary into two disjoint polygons rather than leaving one polygon with a hole. Test 769 (a free ID in the mp-geom bounding box (7.9 1.6, 8.0 1.7)) adds that case.

This mirrors geometry_multipolygon/8 from nimix/osm_conv_tests, a 2012, unlicensed predecessor test suite for the same kind of OSM-to-GIS geometry conversion. We came across it while investigating a geometry bug of our own and wondered whether it had anything grid doesn't; as far as we could tell, this two-point-touch case was the only thing it covered that grid didn't already have. We asked about the relationship between the two repos in nimix/osm_conv_tests#2 — linking it here too since that's the same @joto who'll see this PR.

Both member ways are untagged, matching the plain/no-frills style of nearby fixtures like 706. data.osm/test.json/result/nodes.wkt/ways.wkt follow the existing 7xx ID/coordinate conventions from grid/README.md (nodes 769000-769007, ways 769800/769801, relation 769900; bounding box derived from the test number). The expected WKT was verified against georust/geo's BooleanOps-based even-odd containment resolution, not hand-derived.

data/all.osm, data/tests.json, and grid.db regenerated via make data and make grid.

---

This PR (including the test-case design and the analysis above) was generated by Claude Code, sent on behalf of @brawer.

[joto]

Do you have an actual use case for this?

[brawer]

I'm working on a pipeline that (among other things) assembles the geometry of OSM features as OGC Simple Features; to make sure it works, I run unittests against osm-testdata. Somewhat accidentally, I came across https://github.com/nimix/osm_conv_tests which looks like another, second test corpus for the same problem. So I wondered if that contains any corner cases not yet covered by osm-testdata. Mostly not, with one exception, hence this pull request.

[joto]

Okay, so it's not about some actual current problem with Osmium-based software. I am afraid it goes on the low-priority pile then. Will take me a while to check this, because AI code needs extra careful review.

[brawer]

Yeah, I know. I maintain a bunch of open-source projects myself, and it's really painful to review all these AI-generated changes. Not sure if there's a solution to this, though. Anyhow, feel free to ignore, I just tried to improve your test suite, since it's a really useful piece to have. For my own pipeline, I wrote a separate unit test for the corner case this PR (#11) is about, so this isn't blocking me at all in any way if you don't find the time to review it. — Also, thanks so much for everything you're doing for the OSM world!