feat(gos): reattach #518 sampled-insert paths onto GOS (phase 2)

[zzylol]

Phase 2 of reattaching PR #518 (NitroSketch-style SDK-side row-admission sampling) onto the new GOS stack. Phase 1 (sketchlib-go #77) added the GOS-aware sampled primitives; this PR wires them into the ASAPCollector wrapper layer. Base: #525 (GOS CMS, itself on #524 GOS CountSketch).

The gap this closes

Before this PR, a sampled insert (either a live per-row geometric sampler, or an SDK-pre-decided row-admission bitmask via ApplyAdmittedOccurrence) returned before ever reaching the GOS threshold check. A CountSketch/CMS configured with both sample_p<1 (or SDK row-admission) and gos_delta_epsilon>0 would silently lose insert-time detection for every sampled sample — no crash, estimates stayed unbiased (the 1/p correction is intact), but the GOS mechanism itself never fired for that sketch. The design intends these to compose (the water-filling threshold even carries a sampling-noise floor term), so this was a real integration gap, not an intentional exclusion.

Scope

Only the collector-side wrapper composition. The SDK-side machinery that decides admittedRows/sampleP and gets them onto the wire (aggregationrouter.go, policyfingerprint.go, monitor engine changes, the OTel Go SDK, asap-precompute-rs) is #518's own, much larger, separate concern — out of scope here. That machinery can reattach independently once it needs to call these primitives; this PR just makes sure the collector side is ready for it.

Changes

• observation.go: RowSampled/AdmittedRows/SampleP fields on ObservationValue (lifted from feat(sdk): NitroSketch-style SDK-side row-admission sampling #518 unmodified) — the wire the SDK's row-admission decision travels on into the collector.
• countsketch.go: UpdateString's sampler branch now calls UpdateStringSampledPerRowGOS when GOS is active — this also upgrades CountSketch's sampling from the old whole-item Admit()+divide model to per-row NitroSketch admission (per feat(sdk): NitroSketch-style SDK-side row-admission sampling #518's intent: decorrelates row noise), while preserving GOS composition. Heap-msgpack mode stays excluded (DELTA-HEAP isn't GOS-converted), same gosActive condition ComputeDeltaAgainst already checks. New ApplyAdmittedOccurrence + observer wiring for the SDK-row-admission path.
• cms.go: CMS's existing whole-item WithSampleP/admit() sampling already composed correctly with GOS (InsertWithHashGOS calls admit() internally) — no change needed there. New ApplyAdmittedOccurrence (CMS had none before) + observer wiring, mirroring CountSketch's.

Test plan

• TestCountSketch_RowSampledObserve_ComposesWithGOS / TestCMS_RowSampledObserve_ComposesWithGOS: drive Observe with RowSampled=true through both wrappers, assert the wake signal fires and the drained delta is non-empty.
• TestCountSketch_PerRowSampler_ComposesWithGOS: a live WithSampleP geometric sampler still triggers GOS detection through UpdateString directly.
• go build/go vet/go test clean in asap-precompute-go (full suite, no regressions) and the processor package.

🤖 Generated with Claude Code