Surface a "captures have no session, click to regroup" affordance to users

[mihow]

Summary

Users have no way to tell from the UI that captures in their deployment aren't grouped into sessions. The state manifests as occurrences without a session, stale taxa counts on the project, and no actionable hint — triage today requires Django shell access or the Django admin (only available to staff). We already have a clean pattern for this: the "capture set is empty — click Populate" card. Let's do the same for ungrouped captures.

Observed on a production deployment (2026-04-23): one camera had ~165k (~25%) of its source images with event = NULL, and a corresponding ~70% of that project's occurrences showed "no session" in the UI. From the user's perspective, nothing looked broken — the data was just quietly wrong.

The ask

• Expose an orphan count on Deployment (e.g. source_images_without_event_count) updated by Deployment.update_calculated_fields, same pattern as the existing capture-set size counters.
• When > 0, surface a suggested action card on the deployment page (and bubble up to the project page), matching the "This capture set is empty — click Populate" pattern. Wording: "N captures aren't grouped into a session. Click Regroup to fix."
• The action fires regrouping as a Job (not a bare celery task), so progress / cancel / retry / logs are visible in the jobs UI like every other user-triggered operation. This is what Expose session regrouping as a job stage when syncing images #1157 already proposes; this ticket is about adding the upstream UI that dispatches it.
• On the occurrences list page, if the current filter includes null-session occurrences, show a non-blocking banner linking to the deployment(s) responsible.

Why a separate ticket

The surrounding "make regrouping reliable and visible" problem is already tracked and being worked on — I'm not duplicating that work:

• Fix incorrect session grouping  #904 (OPEN PR) — refactors group_images_into_events with a use_existing flag so incremental regroups after a sync don't redo everything, and adds admin "Fix Sessions" / "Remove from Event" actions. This is the "don't regroup unnecessarily" piece.
• Some sessions are incorrectly grouped #237 (OPEN) — the original incorrect-grouping bug that Fix incorrect session grouping  #904 closes.
• Expose session regrouping as a job stage when syncing images #1157 (OPEN) — expose session regrouping as a job stage in the sync job (and warn about ungrouped captures from e.g. unparseable timestamps). This is the "fail loudly / be visible" piece.
• Make max_time_gap a configurable Project setting #906 (OPEN) — make max_time_gap a per-project setting.
• Capture set population fails with stored kwargs: sample_interval() rejects max_num #1266 (this repo) — unrelated but same class of "job silently leaves DB in partial state, UI shows no signal."

What none of those tickets cover is the user-facing fix-me affordance: once the data is orphaned (either from a pre-#904 sync, a partial sync, or any future corner case), a non-staff user has no way to notice or self-serve a fix. That's the gap this ticket is asking to close.

Acceptance

• Deployment page shows a dismissable/actionable card when orphan count > 0, identical in shape to the capture-set "empty — populate" card.
• Clicking it creates a visible Job (per Expose session regrouping as a job stage when syncing images #1157's job-type work) that users can watch succeed or fail.
• Orphan count goes to zero, card disappears on refresh.
• Occurrence list banner appears when the current filter has null-session results.