[Story]: Define a vendor-neutral AI code-suggestion acceptance schema

## Summary
Define a canonical, vendor-neutral schema for AI code-suggestion acceptance events: `shown / accepted / rejected / edited-before-accept / retained-after-N-days`. Every downstream connector and dashboard depends on this schema existing first. Must explicitly separate ghost-text tab-acceptance (Copilot/Cursor inline completions) from agent-mode diff-approval (Claude Code, Junie-style multi-file proposals) — these are not comparable events and conflating them would corrupt every aggregate metric downstream.

## Intent / Source of truth
Epic finding: "There is no standard to lean on — we must define the schema" (OTel GenAI semantic conventions, `Status: Development`, define token usage/duration/spans but nothing for acceptance). 2026-08-13 backlog consolidation.

## Scope
- [ ] Canonical event schema (fields, enums, timestamps) covering `shown/accepted/rejected/edited-before-accept/retained-after-N-days`
- [ ] Explicit separation of ghost-text tab-acceptance vs agent-mode diff-approval as distinct event subtypes with documented denominator implications
- [ ] Per-vendor denominator mapping doc (what each vendor actually counts as the numerator/denominator for "acceptance rate")
- [ ] Forward-compatibility note: how this schema would remap onto OTel GenAI semantic conventions if/when that spec ever standardizes acceptance

## Out of scope
- Building any connector against this schema (that's stories 2-5)
- Retroactively normalizing historical data collected before this schema existed

## Verification
Schema document reviewed and merged; at least one downstream connector story (2-5) can cite concrete field mappings into this schema without needing schema changes.

## Risk assessment
Getting this wrong early is expensive — every connector built against it inherits the mistake. Prioritize correctness over speed; get the ghost-text/agent-mode separation right before any connector ships.

## AI Usage Declaration
Drafted with AI assistance during the 2026-08-13 cross-repo backlog consolidation. A human owns intent, verification and consequences; acceptance criteria must be verified against the live system before this is closed.