🚸(backend) sort user search results by proximity with the active user (#1802)

## Purpose
Allows a user to find more easily the other users they search, with the
following order of priority:
- users they already share documents with (more recent first)
- users that share the same full email domain
- ~~users that share the same partial email domain (last two parts)~~
- ~~other users~~

Edit: We need to ilter out other users in order to not reveal email
addresses from members of other organisations. It's still possible to
invite them by email.

Solves #1521

## Proposal
- [x] Add a new function in `core/utils.py`:
`users_sharing_documents_with()`
- [x] Use it as a key to sort the results of a basic user search
- [x] Filter user results to avoid reveal of users (and email addresses)
of other orgs or that have not been interacted with.
- [x] User research through "full" email address (contains the '@') is
left unaffected.

---------

Co-authored-by: Anthony LC <anthony.le-courric@mail.numerique.gouv.fr>

src/backend/core/api/viewsets.py

@@ -37,6 +37,7 @@ from csp.constants import NONE
 from csp.decorators import csp_update
 from lasuite.malware_detection import malware_detection
 from lasuite.oidc_login.decorators import refresh_oidc_access_token
+from lasuite.tools.email import get_domain_from_email
 from rest_framework import filters, status, viewsets
 from rest_framework import response as drf_response
 from rest_framework.permissions import AllowAny
@@ -61,7 +62,11 @@ from core.services.search_indexers import (
     get_visited_document_ids_of,
 )
 from core.tasks.mail import send_ask_for_access_mail
-from core.utils import extract_attachments, filter_descendants
+from core.utils import (
+    extract_attachments,
+    filter_descendants,
+    users_sharing_documents_with,
+)
 
 from . import permissions, serializers, utils
 from .filters import DocumentFilter, ListDocumentFilter, UserSearchFilter
@@ -220,18 +225,80 @@ class UserViewSet(
 
         # Use trigram similarity for non-email-like queries
         # For performance reasons we filter first by similarity, which relies on an
-        # index, then only calculate precise similarity scores for sorting purposes
+        # index, then only calculate precise similarity scores for sorting purposes.
+        #
+        # Additionally results are reordered to prefer users "closer" to the current
+        # user: users they recently shared documents with, then same email domain.
+        # To achieve that without complex SQL, we build a proximity score in Python
+        # and return the top N results.
+        # For security results, users that match neither of these proximity criteria
+        # are not returned at all, to prevent email enumeration.
+        current_user = self.request.user
+        shared_map = users_sharing_documents_with(current_user)
 
-        return (
+        user_email_domain = get_domain_from_email(current_user.email) or ""
+
+        candidates = list(
             queryset.annotate(
                 sim_email=TrigramSimilarity("email", query),
                 sim_name=TrigramSimilarity("full_name", query),
             )
             .annotate(similarity=Greatest("sim_email", "sim_name"))
             .filter(similarity__gt=0.2)
-            .order_by("-similarity")[: settings.API_USERS_LIST_LIMIT]
+            .order_by("-similarity")
         )
 
+        # Keep only users that either share documents with the current user
+        # or have an email with the same domain as the current user.
+        filtered_candidates = []
+        for u in candidates:
+            candidate_domain = get_domain_from_email(u.email) or ""
+            if shared_map.get(u.id) or (
+                user_email_domain and candidate_domain == user_email_domain
+            ):
+                filtered_candidates.append(u)
+
+        candidates = filtered_candidates
+
+        # Build ordering key for each candidate
+        def _sort_key(u):
+            # shared priority: most recent first
+            # Use shared_last_at timestamp numeric for secondary ordering when shared.
+            shared_last_at = shared_map.get(u.id)
+            if shared_last_at:
+                is_shared = 1
+                shared_score = int(shared_last_at.timestamp())
+            else:
+                is_shared = 0
+                shared_score = 0
+
+            # domain proximity
+            candidate_email_domain = get_domain_from_email(u.email) or ""
+
+            same_full_domain = (
+                1
+                if candidate_email_domain
+                and candidate_email_domain == user_email_domain
+                else 0
+            )
+
+            # similarity fallback
+            sim = getattr(u, "similarity", 0) or 0
+
+            return (
+                is_shared,
+                shared_score,
+                same_full_domain,
+                sim,
+            )
+
+        # Sort candidates by the key descending and return top N as a queryset-like
+        # list. Keep return type consistent with previous behavior (QuerySet slice
+        # was returned) by returning a list of model instances.
+        candidates.sort(key=_sort_key, reverse=True)
+
+        return candidates[: settings.API_USERS_LIST_LIMIT]
+
     @drf.decorators.action(
         detail=False,
         methods=["get"],
@@ -2338,6 +2405,7 @@ class ConfigView(drf.views.APIView):
         """
         array_settings = [
             "AI_FEATURE_ENABLED",
+            "API_USERS_SEARCH_QUERY_MIN_LENGTH",
             "COLLABORATION_WS_URL",
             "COLLABORATION_WS_NOT_CONNECTED_READY_ONLY",
             "CONVERSION_FILE_EXTENSIONS_ALLOWED",