Phrases documentation for threshold argument is misleading

[umangv]

https://github.com/RaRe-Technologies/gensim/blob/37e49971efa74310b300468a5b3cf531319c6536/gensim/models/phrases.py#L252-L255

It feels to me like this should have said the opposite: "Heavily depends on concrete scoring-function" rather than "Hardly depends on concrete socring-function".

For example, if you choose npmi instead of the default, the threshold has to between -1 and 1, which makes the default (10.0) make no sense. The current documentation suggests that 10.0 will be OK.

[gojomo]

I suspect the intended meaning was, as you surmise, "heavily depends"... and the docs around the 'npmi' option should make that acceptable range clear.

[jenishah]

Hardly has already been replaced by Heavily. Would it be a good idea to add the range of the score under returns in the documentation of the scoring function?