docs/design: import-layer plan + modular-backends reconciliation

[vynulldev]

What & why

Documents the modular-import direction that shipped in the import-layer refactor, and keeps the two design docs from drifting.

• Adds docs/design/import-layer.md: the design for generalizing library import behind an ImportBundle plus an Importer registry. Records why the seam lives in library rather than core (import populates the concrete library.Library and carries rekordbox-shaped side-data the neutral core.Importer never modelled), the decisions (rename MasterDB{Cue,Asset} to Imported{Cue,Asset}, extract the apply pipeline, delete the dead core.Importer), and the phased build-green plan.
• Updates docs/design/modular-backends.md: a Status section plus per-step tags so it agrees with the above. Steps 1-2 shipped; the neutral-cache flip (3d) and the serve-side decoupling (steps 3-4) are deferred until a second protocol can be validated on Denon/StageLinq hardware; step 5 (library formats) was pulled forward and is refined by import-layer.md, which lands the adapter in library (not core) and supersedes the thin core.Importer from step 1.

Docs only — this captures decisions from already-merged work; no code changes.

Hardware testing

• Tested on: N/A: documentation only. No code, protocol, or deck-facing change.

Checklist

• go build ./..., go vet ./..., and go test ./... pass — n/a, docs only (no code changed)
• gofmt -l . is clean — n/a, no Go files
• New source files carry an SPDX header (GPL-3.0-or-later) — n/a, design docs (Markdown) follow the existing docs/design convention of no SPDX header
• Tested on real hardware (deck + firmware noted above), or this change doesn't affect deck behaviour
• I agree my contribution is licensed under the project's GPLv3