fix: key per-profile runtime state by name, not slice index
index shifts when profiles are
reordered/added/removed while the app is running, causing state to attach to the wrong profile.
also i merged repoStatus + last-backup query into one restic call (snapshots --latest 1) when schedule_display is "last", cutting redundant round-trip on startup for slow backends. (for example, i was having this issue with my rclone onedrive backend taking forever to query and sometimes failing even)