fix: drop explicit TokenType::BYTE for OpenFst compatibility (#359)

Both StringCompiler and StringPrinter default to BYTE mode, so the
argument can be omitted entirely. This avoids the TokenType vs
StringTokenType naming difference between the csukuangfj fork and
official OpenFst releases.

Author: pengzhendong

runtime/processor/wetext_processor.cc

@@ -14,15 +14,13 @@
 
 #include "processor/wetext_processor.h"
 
-using fst::TokenType;
-
 namespace wetext {
 Processor::Processor(const std::string& tagger_path,
                      const std::string& verbalizer_path) {
   tagger_.reset(StdVectorFst::Read(tagger_path));
   verbalizer_.reset(StdVectorFst::Read(verbalizer_path));
-  compiler_ = std::make_shared<StringCompiler<StdArc>>(TokenType::BYTE);
-  printer_ = std::make_shared<StringPrinter<StdArc>>(TokenType::BYTE);
+  compiler_ = std::make_shared<StringCompiler<StdArc>>();
+  printer_ = std::make_shared<StringPrinter<StdArc>>();
 
   if (tagger_path.find("zh_tn_") != tagger_path.npos) {
     parse_type_ = ParseType::kZH_TN;