Fix "ignore-case" behavior (facebookresearch#339)

Summary:
Currently, if `ignore-case` is set, the same line will be yielded twice - once as lower-cased version, once as original version, leading to lower than expected uncased scores.
Pull Request resolved: facebookresearch#339

Differential Revision: D12890386

Pulled By: myleott

fbshipit-source-id: 0570e5f6e8f848f2c6439d615e70aca6df097eef

score.py

@@ -44,7 +44,8 @@ def readlines(fd):
         for line in fd.readlines():
             if args.ignore_case:
                 yield line.lower()
-            yield line
+            else:     
+                yield line
 
     def score(fdsys):
         with open(args.ref) as fdref: