flip overlapPointDT[i] / [,j] default: ids = TRUE → FALSE

[jiajic]

Flip the default for `[` on `overlapPointDT` single-axis indexing from `ids = TRUE` (return integer indices) to `ids = FALSE` (return subset object). The selection-query form is still available via `ovlp[i, ids = TRUE]`.

Why

The previous default was the unusual case. Most callers want subset semantics:

• `.subset_overlaps_poly` used `[i, ids = FALSE]` to work around the default
• `.subset_overlaps_feat` used `[, i, ids = FALSE]` for the same reason
• `overlapIntensityDT[i]` has no `ids` param — always returns subset
• `SpatVector[i]` (legacy storage) also returns subset
• The "return indices" default was specific to `overlapPointDT`, creating an asymmetry across the three classes that store overlap data

Audit found zero external callers (Giotto, GiottoDisk, GiottoVisuals) relying on the old default — only the test in `test-aggregate.R` and the documented examples.

What changes

• Two `setMethod("[", "overlapPointDT", ...)` defaults flipped + branch order inverted for readability (explicit `ids = TRUE` selects, default subsets).
• `.subset_overlaps_poly` / `.subset_overlaps_feat` simplified to `[i]` / `[, i]` — the `ids = FALSE` workaround is no longer needed.
• `test-aggregate.R`: the selection-query tests now pass `ids = TRUE` explicitly; new assertions cover the default subset behavior.
• Class docs updated to reflect the new default.

Side benefit

Standalone polygon loads (e.g. `GiottoData::loadSubObjectMini`) that carry legacy `SpatVector` overlaps now subset cleanly. `SpatVector[i]` returns a subset `SpatVector`, which is what `.subset_overlaps_poly` now expects. No migration hook required.

Test plan

• Full test suite: 970 / 0.
• Standalone-loaded polygon with legacy SpatVector overlaps subsets correctly (`GiottoData::loadSubObjectMini("giottoPolygon")` + `gpoly[ids]` + `gpoly[c(T,F)]`).
• Both forms still work via tests: `ovlp[i]` (default, subset) and `ovlp[i, ids = TRUE]` (selection, integer vector).

🤖 Generated with Claude Code