An endpoint outside the frustum is not sampled and projects at its 0 m
stand-in height. The horizon test then judged that stand-in, so an arrow
whose far end lies beyond the sea-level horizon but above it at its real
height disappeared with its label. Such an endpoint now skips the point
horizon test; the height-range test still drops endpoints that are
provably hidden, and settled or in-frustum anchors keep the point test.
Adds a regression with a real frustum and occluder (camera 100 m up, far
end 56 km away on 100 m terrain). CURRENT-STATE and the changelog now say
the 32-sample budget applies per projection pass, and the redundant helper
comment in the test is gone.
The last merge of main kept four tests on a DOM harness that main had
removed, so they threw ReferenceError and CI failed. They now run on the
store-backed board. The merge had also dropped main's empty-board publish,
the leaving/byKey clear in destroy, and three unrelated CURRENT-STATE
paragraphs; all are restored. The projection now builds each height-cache
key once per point, and the changelog entry says what changed.