fix(statistics): use subquery instead of join to avoid cartesian product

The way we filter in the statistics view, any added filter (that affects
reports) adds a "dimension" to the cartesian product, exploding the total
number of hours reported.

Instead of using JOIN, we do EXISTS(SUBQUERY) now, which should avoid this
issue. Might be a tiny bit slower, but let's try to make it correct first, then fast.

backend/timed/reports/views.py

@@ -8,7 +8,7 @@
 from zipfile import ZipFile
 
 from django.conf import settings
-from django.db.models import F, Q, QuerySet, Sum
+from django.db.models import F, Q, Exists, OuterRef, QuerySet, Sum
 from django.db.models.functions import ExtractMonth, ExtractYear
 from django.http import HttpResponse
 from django.utils.http import content_disposition_header
@@ -117,9 +117,14 @@ def filter(self, /, **kwargs):
         return new_qs
 
     def filter_base(self, *args, **kwargs):
+        filtered = (
+            self.model.objects.filter(*args, **kwargs)
+            .values("pk")
+            .filter(pk=OuterRef("pk"))
+        )
         return StatisticQueryset(
             model=self.model,
-            base_qs=self._base.filter(*args, **kwargs),
+            base_qs=self._base.filter(Exists(filtered)),
             catch_prefixes=self._catch_prefixes,
             agg_filters=self._agg_filters,
         )