landing: central container memory/load stats for all three sites

Collect RSS groups, top processes, and loadavg from inside each podman
container (not host /proc), merge into performance.memory, show compact
labels with cgroup limits and byte tooltips, and cover this in the test suite.

scripts/taler-landing/collect_bank_stats.py

@@ -629,13 +629,9 @@ def main() -> int:
         if int_http != "200":
             int_ms, int_http = measure_ms(f"{bank}/taler-integration/config")
 
+        # loadavg filled later by host merge from *inside* the bank container
+        # (host /proc would be wrong when this script runs on koopa outside podman)
         loadavg = ""
-        try:
-            with open("/proc/loadavg", encoding="utf-8") as f:
-                parts = f.read().split()
-                loadavg = ",".join(parts[:3])
-        except Exception:
-            pass
 
         total_pack = pack_amt(total_wd)
 
@@ -755,7 +751,8 @@ def main() -> int:
                 "loadavg": loadavg,
                 "memory": {
                     "container_rss_human": "—",
-                    "note": "memory filled by host collector merge when available",
+                    "container_rss_label": "—",
+                    "note": "filled by collect_container_resources.sh (in-container /proc + cgroup)",
                 },
             },
             "alt_unit_names": alt,