Adding temperature scaling on Joiner logits: (#789)

* Adding temperature scaling on Joiner logits:

- T hard-coded to 2.0
- so far best result NCE 0.122 (still not so high)
    - the BPE scores were rescaled with 0.2 (but then also incorrect words
      get high confidence, visually reasonable histograms are for 0.5 scale)
    - BPE->WORD score merging done by min(.) function
      (tried also prob-product, and also arithmetic, geometric, harmonic mean)

- without temperature scaling (i.e. scale 1.0), the best NCE was 0.032 (here product merging was best)

Results seem consistent with: https://arxiv.org/abs/2110.15222

Everything tuned on a very-small set of 100 sentences with 813 words and 10.2% WER, a Czech model.

I also experimented with blank posteriors mixed into the BPE confidences,
but no NCE improvement found, so not pushing that.

Temperature scling added also to the Greedy search confidences.

* making `temperature_scale` configurable from outside

sherpa-onnx/csrc/online-recognizer.cc

@@ -96,6 +96,8 @@ void OnlineRecognizerConfig::Register(ParseOptions *po) {
   po->Register("decoding-method", &decoding_method,
                "decoding method,"
                "now support greedy_search and modified_beam_search.");
+  po->Register("temperature-scale", &temperature_scale,
+               "Temperature scale for confidence computation in decoding.");
 }
 
 bool OnlineRecognizerConfig::Validate() const {
@@ -142,7 +144,8 @@ std::string OnlineRecognizerConfig::ToString() const {
   os << "hotwords_score=" << hotwords_score << ", ";
   os << "hotwords_file=\"" << hotwords_file << "\", ";
   os << "decoding_method=\"" << decoding_method << "\", ";
-  os << "blank_penalty=" << blank_penalty << ")";
+  os << "blank_penalty=" << blank_penalty << ", ";
+  os << "temperature_scale=" << temperature_scale << ")";
 
   return os.str();
 }