Build the language model based on n-grams frequencies

[iosadchiy]

This is not intended to be merged but to discuss if this feature can be useful.

I was experimenting with n-grams frequencies from Ruscorpora. The idea was to load the frequency files directly into the model:

jamspell load_ngrams alphabet_ru.txt 1grams.csv 2grams.csv 3grams.csv ru.bin

You can see some short samples of the .csv files here

Let me know if this feature can be useful.

[bakwc]

Thanks for PR, good feature! Let me know when you finish - I'll be glad to merge it. BTW, could you please upload somewhere your model? I'd like to compare it to a model trained on wikipedia+news.

[iosadchiy]

Yep, sure, the model is here

[rprilepskiy]

@iosadchiy, did you have time to fix the PR checks error?