New config parameter to choose SentencePiece unicode normalization

Author: ZJaume

configs/autogenerated/en-pl-example.yml

@@ -36,6 +36,7 @@ experiment:
   spm-sample-size: 10_000_000
   spm-vocab-size: 32000
   spm-vocab-split: false
+  spm-norm-rule: nmt_nfkc
   teacher-ensemble: 1
   teacher-mode: two-stage
   teacher-decoder: ctranslate2