Web session list loads slowly on refresh — listArtifacts scans archived sessions too

[pure-craft]

Symptom

After refreshing the DSH Web page, the workspace session list in the sidebar takes a noticeable delay (several seconds) to appear.

Investigation

I traced the session list loading pipeline:

1. Web refresh → SessionManager.handleConnected() → refreshList() → remote.session.list({})
2. Host-side ApiSessionList.list() → sessionQuery.listSessions() → persistence.list()
3. JsonlSessionPersistence.listArtifacts() iterates over all project directories and session directories, performing open → read → Zstd decompress → JSON parse header for every session file on disk.

The key finding: listArtifacts() scans every session file indiscriminately, including archived ones. The archive filter (archivedSessionIds) is applied only in the client-side UI layer (sessionVisible()), well after the full I/O cost has already been paid.

Why this matters

• As a user accumulates sessions over time, the list() I/O cost grows linearly with the total number of persisted sessions — not just the active ones.
• Archiving a session is meant to "hide it out of the way," but archived sessions still participate in every list scan. Archiving currently provides zero performance benefit.
• Even when a user only cares about a handful of active sessions in the current workspace, they still pay for reading headers of every historical session across all workspaces.

Possible improvements

1. Let listArtifacts() or list() accept an exclude set so archived session IDs can be skipped at the file-scanning stage (this would require coordination between workspaceRegistry and sessionPersistence).
2. Provide a lighter readdir-only fast path that returns directory names / session IDs without reading headers, letting the client lazy-load metadata on demand.