fix(cortex): initial ingest tolerates a missing/unreadable vault root

[ApiliumDevTeam]

Problem

On macOS, after installing an in-app Akashi update and restarting, the engine died with:

Engine unavailable — initial ingest failed: Internal error: walk error: … /…/akashi/vaults/<hash>/ — No such file or directory (os error 2)

The post-update restart races with vault working-copy setup (or an iCloud vault hasn't synced down), so the ingest root doesn't exist yet. The ignore walker yields an Err entry that ? turned into a fatal Error::Internal, taking the whole engine down.

Fix

aingle_cortex::service::ingest::ingest_path is now resilient:

• Missing root → early-return an empty IngestReport with a warn! (engine stays up; a later ingest picks up files once the vault is mounted).
• Per-entry walk errors and per-file read errors → warn! + continue instead of ? (tolerates iCloud .icloud placeholders and permission errors).

Regression test ingest_missing_path_is_graceful reproduces the crash and passes. All 6 ingest tests green; cargo fmt applied; file is clippy-clean.

Akashi consumes this via a path dep (akashi-core/engine.rs initial_ingest), and its release CI checks out aingle at main, so merging here ships the fix to the next Akashi build.