fix(ecstore): terminate walk directory streams (#6462)

* fix(ecstore): terminate walk directory streams

* test(e2e): refresh node service selection

* fix(filemeta): initialize empty metacache streams

.config/e2e-full-selection.txt

@@ -1,2 +1,2 @@
-sha256-darwin=f832043fcca8c0b616c5d820a3a652da7544298ef5812a8668a3a9a3e4607b8b
-sha256-linux=93b94adb110b86a41d0b7313909e0bf53cb1515e2d08e8f105652b29b249990f
+sha256-darwin=b63c54946d978e22660762e8e0351f6839bebf9be2282fe08e03e775e4303f0a
+sha256-linux=ef3be856bd3257c2c369428f66a48ad073dad8187229fe95840a600a81edf22b

.config/nextest.toml

@@ -315,7 +315,7 @@ slow-timeout = { period = "60s", terminate-after = 2, grace-period = "10s" }
 #     the target incl. multipart and the resync path, SSE-C and
 #     target-without-KMS stay fail-closed), and one guards event/history
 #     observers.
-#   * 12 `_real_dual_node` site-replication tests — each spawns TWO full rustfs
+#   * 13 `_real_dual_node` site-replication tests — each spawns TWO full rustfs
 #     servers and drives the cross-process site-replication control plane.
 #   * 1 `_real_three_node` site-replication test.
 #   * 1 `_real_single_node` service-account round-trip test.
@@ -406,7 +406,7 @@ path = "junit.xml"
 #     object_lambda) — too heavy for the merge budget; they run in the
 #     e2e-nightly serial cluster-fault lane.
 #   * replication_extension_test — repl-1 already splits it into the PR
-#     `e2e-smoke` (20 fast) and `e2e-repl-nightly` (55 slow) lanes and reserves
+#     `e2e-smoke` (20 fast) and `e2e-repl-nightly` (56 slow) lanes and reserves
 #     it for those, so e2e-full does not double-run it.
 #   * #[ignore]d tests — nextest skips them by default (no --run-ignored); the
 #     manual-localhost:9000 reliant/policy tests are ci-13's migration.