fix(tests): do not look into full path for error metric

In order to collect all error metrics we're checking for "err" in the
key name, which causes performance tests to fail because we added
metrics called "intERRupts". Change the check to only look at the leaf
of the path, so that when we have a metric such as
`fc_metrics.net_metrics.cfg_fails`, we'll only be looking into
`cfg_fails`.

Suggested-by: Riccardo Mancini <[email protected]>
Signed-off-by: Babis Chalios <[email protected]>

Author: bchalios

tests/host_tools/fcmetrics.py

@@ -479,7 +479,10 @@ def flatten_dict(node, prefix: str):
     failure_metrics = {
         key: value
         for key, value in flattened_metrics.items()
-        if "err" in key or "fail" in key or "panic" in key or "num_faults" in key
+        if "err" in key.split(".")[-1]
+        or "fail" in key.split(".")[-1]
+        or "panic" in key.split(".")[-1]
+        or "num_faults" in key.split(".")[-1]
         if value
         if key not in ignored_failure_metrics
     }