Feed the fast parser into typed rollout documents

[iam-brain]

Question

Add an isolated adapter PR that converts output from the existing cancellation-aware fast rollout parser—metadata plus complete token snapshots—into typed lineage-ledger Document values without duplicating JSON parsing or changing published totals.

[iam-brain]

Resolved by preparing and validating Feature/Fix: Adapt rollout snapshots for lineage accounting. The existing fast JSONL pass now emits typed complete observations and lineage identity metadata without changing published totals.