Fix bug

Results from python3 src/python/localrun.py -source combinedFieldsUnevenlyWeightedBig:

Run 1:
	    TaskQPS baseline      StdDevQPS my_modified_version      StdDev                Pct diff p-value
     CFQHighHigh        2.81     (10.9%)        2.40     (10.3%)  -14.6% ( -32% -    7%) 0.000
      CFQHighMed        3.27     (11.8%)        2.91     (11.0%)  -11.2% ( -30% -   13%) 0.002
      CFQHighLow       16.09     (14.6%)       14.58     (11.9%)   -9.4% ( -31% -   20%) 0.025
	PKLookup       92.08     (11.6%)       96.23     (13.3%)    4.5% ( -18% -   33%) 0.255

Run 2:
	    TaskQPS baseline      StdDevQPS my_modified_version      StdDev                Pct diff p-value
      CFQHighLow       15.76     (12.0%)       10.19     (12.9%)  -35.3% ( -53% -  -11%) 0.000
      CFQHighMed       12.01     (13.3%)        8.79     (13.9%)  -26.9% ( -47% -    0%) 0.000
     CFQHighHigh        5.59     (10.8%)        4.40      (8.4%)  -21.2% ( -36% -   -2%) 0.000
	PKLookup       93.36     (21.9%)       94.29     (22.6%)    1.0% ( -35% -   58%) 0.888

lucene/sandbox/src/java/org/apache/lucene/sandbox/search/CombinedFieldQuery.java

@@ -749,7 +749,7 @@ private List<Impact> mergeImpactsAcrossFields(
             // computation
             // on merged impacts across fields
             long maxFreqSum = 0;
-            long minNorm = Long.MIN_VALUE;
+            long minNorm = Long.MAX_VALUE;
             for (List<Impact> impacts : mergedImpactsPerField.values()) {
               // highest freq at the end of each impact list
               maxFreqSum += impacts.get(impacts.size() - 1).freq;