fix langdata (user words/patterns) file suffixes for LSTMs:

[bertsky]

• add another constructor for LSTMRecognizer
  which takes the language_data_path_prefix configured/selected
  at runtime and passes it to the internal CCUtil
• use this in Tesseract::init_tesseract_lang_data when LSTMs
  are available

(this was missing from 297d7d8)

[bertsky]

Fixes #2584

[bertsky]

But maybe anyone has a better idea of how to make this work with LSTMs?

(The story is the same as for the configuration variables in #2328 – we have to get the prefix into the shallow CCUtil / Dict inside LSTMRecognizer. For the config variables I decided to pass the existing ParamsVectors via LSTMRecognizer::LoadDictionary(), resetting the defaults only for the specific variables to be shared, via Param::ResetFrom(). But here we are dealing with the combined runtime settings for tessdata prefix and language. We could alternatively use TessdataManager for this, maybe by giving it an additional member language_data_path_prefix_ to be set in the constructor. I even thought about using its existing member data_file_name_ and just truncate the .traineddata suffix, but I guess this is bound to fail under certain conditions I am unable to identify.)

[zdenop]

thanks.