Parity: cite the generated surface report instead of re-asserting it

[mgravell]

Follow-up to #212, which made "what does Dapper.AOT do with this overload?" a derived fact. This stops parity.md answering that question from memory.

• ApiSurface.expected.txt is named as the source of truth for API-surface status, with the three things it deliberately does not cover spelled out: behaviour, per-call-site refusals, and non-extension statics (Format, PackListParameters, LookupDbType, TypeHandlerCache<T> — called directly rather than intercepted, so the generator has no opinion and nothing measures whether they are AOT-safe when called).
• Three ❓ rows are settled by the report and now cite it: Parse (all three overloads sit outside the generator's name filter — emitted nothing, said nothing), Format/ReplaceLiterals (same, and Format is not even an extension method), and CommandDefinition.
• CommandDefinition was the worst of the stale rows. It read as an unverified question about CommandFlags.Pipelined. It is in fact 27 overloads, every one skipped in silence — high impact, not the "med" it carried. Split into the overload row and a separate CommandFlags row, since the latter is moot at a call-site until the former is fixed.
• The delta summary leads with both measurements, because they answer different questions: 40 of 110 overloads tell the consumer nothing, and 677 of 793 corpus tests pass.
• Adds a row 0 — say something at those 40 mute overloads. It supports no new API and is small, but it converts a silent runtime AOT failure into a build warning, which makes it the cheapest safety item on the list.

Remaining ❓ now means "needs a corpus run to settle" unless a row says otherwise; the few things nothing measures are marked uninstrumented in place.

Notes only — no code changes.