Include recommend_for_all_users functionaliy

[jerrytim]

it's really helpful to have this function to recommend top N items to all users. N is the function argument. For each user, tuple (itemid, rating) is returned.
It's similar to Spark ALS model.

[benfred]

I agree that this is useful functionality - the only tricky bit is doing this efficiently in parallel (rather serially like I did here: https://github.com/benfred/implicit/blob/master/examples/lastfm.py#L122).

Going to include this when as part of the evaluation code

[benfred]

the latest version (v0.3.9) has a 'recommend_all' for MF models that achieves this

[jerrytim]

Is there a way to get the scores as well in this recommend_all() function? @benfred