cleanup: drop unused LINK_ISOLATION hack

This conditional was added as an escape hatch just in case the bug fix
for RHELDST-21893 broke some exodus-gw users and would have to be
temporarily disabled. In practice we never needed to use this and it can
be cleaned up. Link isolation should always be enabled, as originally
intended.

exodus_gw/models/publish.py

@@ -22,22 +22,6 @@
 
 from .base import Base
 
-# The resolution of links SHOULD be isolated to within the current publish;
-# this is very important. However, the original implementation of the code
-# did not do that and allowed links to be resolved across any publish,
-# see RHELDST-21893.
-#
-# That is being fixed, but the problem is that clients may have come to
-# rely on it. In particular, if pub/pulp/exodus integration is *disabled*,
-# we rely on it.
-#
-# Thus, this semi-hidden setting acts as an escape hatch to re-enable
-# the old buggy behavior if it turns out to be needed. Not a proper setting.
-#
-# Only set this if you really know it's needed! With any luck, this branch
-# should be removed quickly.
-LINK_ISOLATION = (os.environ.get("EXODUS_GW_LINK_ISOLATION") or "1") == "1"
-
 
 class Publish(Base):
     __tablename__ = "publishes"
@@ -108,11 +92,9 @@ def resolve_links(
         match_bundle: Bundle[Any] = Bundle(
             "match", Item.web_uri, Item.object_key, Item.content_type
         )
-        query = db.query(match_bundle).filter(Item.web_uri.in_(ln_item_paths))
-        if LINK_ISOLATION:
-            # See comments above where LINK_ISOLATION is set.
-            # This path should become the only path ASAP!
-            query = query.filter(Item.publish_id == self.id)
+        query = db.query(match_bundle).filter(
+            Item.publish_id == self.id, Item.web_uri.in_(ln_item_paths)
+        )
 
         matches: dict[str, dict[str, Optional[str]]] = {
             row.match.web_uri: {