fix: only use digit-by-digit reading for 4-digit years (#349)

* feat: support nbest output in normalize/tag (#296)

normalize(input, nbest=1) returns a single string (default, backward
compatible). normalize(input, nbest=N) returns a list of N-best results.

Example:
  n.normalize('海淀区108号', nbest=3)
  # ['海淀区一百零八号', '海淀区十八号', '海淀区幺零八号']

* fix: only use digit-by-digit reading for 4-digit years (#302)

Previously 2-digit numbers followed by 年/年度/赛季 were read
digit-by-digit (38年 => 三八年). Now only 4-digit years use this
pattern, so 38年 => 三十八年 (via number) while 2024年 => 二零二四年.

Author: pengzhendong

tn/chinese/rules/measure.py

@@ -47,12 +47,12 @@ def build_tagger(self):
             measure @= self.build_rule(cross("两" + unit, "二" + unit), l="[BOS]")
             measure @= self.build_rule(cross("到两" + unit, "到二" + unit), r="[EOS]")
 
-        # -xxxx年, -xx年
+        # xxxx年, xxxx-xxxx年
         digits = self.cardinal.digits
-        cardinal = digits**2 | digits**4
+        yyyy = digits**4
         unit = accep("年") | accep("年度") | accep("赛季")
-        prefix = cardinal + (rmspace + unit).ques + to
-        annual = prefix.ques + cardinal + unit
+        prefix = yyyy + (rmspace + unit).ques + to
+        annual = prefix.ques + yyyy + unit
 
         tagger = insert('value: "') + (measure | annual) + insert('"')