fix!: the macOS test the Option change broke, and displays in the ontology

CI was red on macOS: `tests/macos_readers.rs` compares swap used against
total and I updated `tests/plausibility.rs` for the `Option` fields and
missed it. It is `cfg(target_os = "macos")`, so the local suite never built
it — and neither did the cross-target checks, because those ran `cargo check`
without `--all-targets` and so skipped every test target.

Both cross-checks now pass `--all-targets`, and HANDOFF records why. A
cross-check that builds only the library is a cross-check that cannot see a
platform-gated test.

Also corrects an overclaim. "Item F is complete for every cluster with a
reader behind it" was wrong: it is complete for every cluster the *plan
named*. Probing the rest found four that answer on this machine and have no
schema at all — displays, input devices, services and storage controllers —
and around 28 modules in that state. HANDOFF now lists them.

Displays are the first of those: `board.display.{n}.*` with a
`board.display.<none>` diagnostic. Verifying it on the machine at hand, as
the standing instruction says, immediately found the reader publishing
`0x0 at 0 Hz` as `measured` for an attached and correctly named LG ultrawide
whose current mode it could not read. A display is not zero pixels wide;
those are now unavailable with the reason.

The conformance suite then caught my own declaration: `connection` was
declared non-nullable and two of three displays report a type the reader
cannot classify. The entity was wrong, not the reader.

821 lib tests, 21 ontology conformance, all three targets clean with
`--all-targets`.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>