[codex] Make signal coverage scalar optional

[bokelley]

Summary

Makes deprecated coverage_percentage optional in get_signals and wholesale signal feed payloads while keeping coverage_forecast authoritative for signal-level coverage. Updates signal docs, compliance storyboards, and schema tests to cover aggregate present buckets, low/high coverage bounds, and wholesale payloads without the legacy scalar.

Validation

Ran npm run test:schemas, npm run test:composed, npm run test:json-schema, npm run build:schemas, npm run build:compliance, precommit checks, pre-push storyboard matrices, and docs broken-link validation. npm run test:snippets is still known-red for unrelated auth/live-fixture drift; tracked separately in #5100.

Closes #5089.

[aao-release-bot]

LGTM. Surgical relaxation of an already-deprecated field; coverage_forecast is the authoritative discovery surface and the scalar stays fallback-only.

Things I checked

• static/schemas/source/signals/get-signals-response.json and static/schemas/source/core/wholesale-feed-event.json: only the required arrays drop coverage_percentage. type: number, minimum: 0, maximum: 100, and deprecated: true are preserved verbatim. Descriptions updated to name coverage_forecast authoritative.
• presence: "present" without signal_value is schema-legal per static/schemas/source/core/forecast-point-dimensions.json — the conditional only constrains the type of signal_value when present, doesn't promote it to required. New test at tests/schema-validation.test.cjs:662-679 pins this shape.
• completeExclusiveCoverageRatesPartition rewrite at tests/schema-validation.test.cjs:787-796 iterates low/mid/high, skips a level if every point omits it, fails if some points have it and others don't. Right invariant.
• Changeset exists at .changeset/signal-coverage-legacy-scalar.md with minor. Defensible — .changeset/4796-review-followups.md is the precedent (dropping a required array entry on a deprecated field shipped as minor). Field has been advertised deprecated: true since #5088 landed in the same cycle, so strict consumers had a window.
• Compliance YAML: signal-marketplace/index.yaml drops the field_present: signals[0].coverage_percentage assertion. signal-owned/index.yaml never had an equivalent assertion (its validations at L134-153 only cover signal_agent_segment_id, pricing_options, context, signal_id.source) — symmetric with marketplace by construction. universal/schema-validation-signals.yaml narrative correctly trimmed to schema-required fields.
• No doc surface still calls coverage_percentage required.

Follow-ups (non-blocking — file as issues)

• Strict-validator adopter-action row. Codegen consumers (Pydantic, quicktype, OpenAPI) will see the field flip required→optional and need fixture refreshes. Worth a row in docs/reference/release-notes.mdx matching the precedent set by #4796.
• Storyboard assertions trail the narrative. signal-marketplace/index.yaml:118 and signal-owned/index.yaml:118 expected: blocks now describe coverage_forecast as the disclosure surface but neither validates signals[0].coverage_forecast presence. The prose moved ahead of the assertions.
• min_coverage_percentage request filter. docs/signals/tasks/get_signals.mdx:711 still documents it as operating on the legacy scalar. With the scalar now optional, the filter's denominator semantics get muddier — a sentence on filter behavior when the scalar is absent would help.

Minor nits (non-blocking)

1. Vacuous-pass on empty partition. tests/schema-validation.test.cjs:787-796 returns true if every point omits all of low/mid/high. Pre-existed in the mid-only version, inherited here. Tracking anyLevelPresent and returning false if no level had numeric data on any point closes the gap.
2. Idiomatic clone. tests/composed-schema-validation.test.cjs:732 builds signalListingCoreWithoutLegacyCoverage via spread+delete; const { coverage_percentage: _legacy, ...rest } = signalListingCore is one line. Pure style.

Approving on the strength of the surgical required-array trim plus the new low/high partition coverage. Safe to merge.