Fix NAN correlation values

Really short regions confuse the correlation computations because the
duration is always zero. In this case, we assign the correlation value 0
because such a short duration is not latency-critical at all.

xlap/analyse/latency.py

@@ -175,7 +175,10 @@ def analyse(df, config):
                 l3 = []
                 for i in range(len(l1)):
                     l3 += [_get_timestamp(df,event2,i) - _get_timestamp(df,event1,i)]
-                correlation = pearsonr(l3, e2e)[0]
+                if any(map(lambda x: x != 0, l3)):
+                    correlation = pearsonr(l3, e2e)[0]
+                else:
+                    correlation = 0
 
                 hdb += [{'Start':str(event1), 'End':str(event2), 'Correlation':correlation}]