Feature: Handling Cases for User to Customer Relationships

This commit handles a case where customer data are "dead-data" menaing there is no way of erasure for them, which makes the app non-compliant

docs/privacy.md

@@ -190,10 +190,71 @@ $service->cancelErasure($request);
 Authentication is never blocked by deactivation — `UserOtpService::validate()` still requires
 the correct OTP code. Once validated, it dispatches `Modules\Core\Auth\Events\UserAuthenticated`;
 `Modules\Core\Privacy\Listeners\CancelErasureOnLoginListener` (registered in
-`PrivacyServiceProvider`) looks for a pending request keyed on *that User's own id* — never a
-Customer-scoped one, since Customer-scope never deactivates a login in the first place — and
-calls `cancelErasure()` on it. Logging back in **is** the "I changed my mind" action — no
-separate UI/flow needed for reactivation.
+`PrivacyServiceProvider`, **queued** — see below) looks for a pending request keyed on *that
+User's own id* — never a Customer-scoped one, since Customer-scope never deactivates a login in
+the first place — and calls `cancelErasure()` on it, then reverts every Customer erasure request
+it caused (see "The sole-owner cascade" below). Logging back in **is** the "I changed my mind"
+action — no separate UI/flow needed for reactivation.
+
+This listener is queued rather than synchronous, so login returns to the browser without waiting
+on the bookkeeping. Nothing else in this codebase currently reads `deactivated_at` besides this
+listener and `PrivacyService` itself — `UserOtpService::validate()` never gates the login on it —
+so the brief window between the login response and the job actually running has no other consumer
+to observe it as stale.
+
+### The sole-owner cascade — erasing the last User on a Customer also erases the Customer
+
+If a User is erased and they were the **only** User linked to a given Customer, that Customer's
+data (orders, addresses, buyer record) becomes permanently unreachable through any login the
+moment the User's identity is gone — nobody could ever again log in to exercise a data-subject
+right over it. GDPR's data minimization principle (Art. 5(1)(c)) means it shouldn't just sit
+there indefinitely with no legitimate purpose.
+
+`requestErasureForUser()` and `requestImmediateErasureForUser()` both fire
+`Modules\Core\Privacy\Events\UserErasureRequested` right after the request is created (and, for
+the immediate path, before `completeErasure()` runs — see below).
+`Modules\Core\Privacy\Listeners\CascadeCustomerErasureListener` (**queued**, registered in
+`PrivacyServiceProvider`) handles it: for every Customer the User is linked to, if that User is
+currently the *sole* linked User (count is 1, and that one User is this one — not just count ===
+1, to be explicit rather than relying on an assumption), it opens a second, independent
+grace-period request via `requestErasureForCustomer($customer, $user, causedByRequestId: ...)`.
+Both requests then run through their own separate 30-day windows.
+
+```
+User erasure requested
+   │
+   ▼
+UserErasureRequested event ──▶ CascadeCustomerErasureListener (queued)
+                                    │
+                                    ▼
+                    for each linked Customer: sole owner?
+                                    │ yes
+                                    ▼
+                    requestErasureForCustomer(..., causedByRequestId: <user request id>)
+```
+
+**Tracing the cascade — `caused_by_request_id`.** A cascade-created Customer request's
+`caused_by_request_id` points back at the User request that triggered it. This is what lets
+`CancelErasureOnLoginListener` revert *exactly* the cascade a User's own cancellation should
+undo (via `DataErasureRequest::caused()`) without ever touching an unrelated, independently
+staff-requested Customer erasure the User happens to still be linked to.
+
+**Why this is queued, not synchronous.** `CascadeCustomerErasureListener` runs as an independent,
+separately-retryable job rather than inline inside `requestErasureForUser()` — a failure in the
+cascade check never rolls back or blocks the User's own request, and there's no
+`DB::transaction()` wrapping needed, since the two writes (the User's request, and any cascaded
+Customer request) aren't required to be atomic with each other.
+
+**A known, accepted race on the immediate-erasure path only.** Because the listener is queued,
+Eloquent re-fetches its models fresh when the job actually runs (see
+`Illuminate\Queue\SerializesModels`) — so `$event->request->subject->customers` reflects the
+*real* state at execution time, not a stale snapshot from dispatch time. For
+`requestImmediateErasureForUser()`, that job may run before or after `completeErasure()` detaches
+the User's memberships in the same call. If the detach happens first, the User is simply no
+longer linked to anything by the time the cascade job runs, and nothing cascades — an accepted
+race for that rare, staff-only path (see "Immediate erasure" below), not a concern for the
+everyday `requestErasureForUser()` grace-period path, where nothing detaches until its own later,
+separate `completeErasure()` run — well after the cascade job has had time to fire.
 
 ### Processing due requests — one job per request
 
@@ -305,7 +366,11 @@ whole point is for these tables to remain readable after the record they're abou
 erased. `DataErasureRequest::isForCustomer()` tells you which scope a given request is.
 
 `DataErasureRequest.requested_by_type`/`requested_by_id` capture who asked for it (the subject
-themselves, self-service, or `Staff` acting on their behalf) at request time.
+themselves, self-service; `Staff` acting on their behalf; or, for a cascade-created Customer
+request, the User whose erasure caused it — see "The sole-owner cascade") at request time.
+`DataErasureRequest.caused_by_request_id` is set only on a cascade-created Customer request,
+pointing back at the User request that triggered it; null on every normal, directly-requested
+erasure — see `DataErasureRequest::causedBy()`/`::caused()`.
 `DataErasureRequest.report` holds the full per-provider outcome once `completeErasure()` runs;
 `DataExportRequest.file_path` points at the generated zip once `WriteExportToCsvListener`
 finishes.