Expand FromRow tuple impls to 16 (currently capped at 4)

Start in hyperdb-api/src/result.rs at the existing FromRow tuple implementations around lines 758-787, and review §5 of docs/RUST_API_GAP_ANALYSIS.md for context. Extend the implementations through 16-tuples with a declarative macro, add real-query sanity tests for the requested arities, and update the trait rustdoc to describe the tuple cap.

Summary

The FromRow blanket tuple impls only cover 1, 2, 3, 4-tuples. Common Rust DB crates (sqlx, rusqlite, tokio-postgres) go to 12 or 16. Rows with more than four columns must define a hand-written struct + FromRow impl even when the user just wants ad-hoc destructuring.

Current state

$ grep -n "FromRow for (Option" hyperdb-api/src/result.rs | wc -l
3

Plus the 1-tuple impl at result.rs:758, giving four total — see result.rs:758-787.

Original gap analysis: §5 of docs/RUST_API_GAP_ANALYSIS.md (predecessor repo).

Proposed work

• Extend the tuple impls up to 16-tuple, matching sqlx. Generate them via a declarative macro to keep the code compact.
• Add a few sanity tests for 5/8/12/16-tuple cases against a real query.
• Document the tuple-cap in the trait rustdoc.

Backwards compatibility

Purely additive. Existing (Option<A>,) through (Option<A>, Option<B>, Option<C>, Option<D>) impls keep working unchanged.

Performance note

Pure type-system addition — zero runtime cost. Compile time grows slightly, but only for translation units that actually use the new arities.

Related

• #61 — #[derive(FromRow)] reduces the demand for high-arity tuple impls (most users will reach for the derive instead). Tuple impls remain useful for ad-hoc destructuring without defining a struct.