Starting on FAQ and Helpful Discussions

[SimonCW]

This PR might evolve into a useful FAQ section. For the moment, my goal is to keep track of interesting discussions while closing issues and start a discussion on how to best document these discussions to provide value for users and avoid unnecessary questions /issues being created.

I see two types of of useful documentation:

1. FAQ

This should include often asked questions that concern the package directly and are straightforward to answer, e.g.:

• "Is there a GPU based option for LightFM?" -> No, there is no option to run training or inference on the GPU with LightFM and there are currently no plans to change this.

2. Useful discussions and strategies (Majority of issues)

This should include often raised question that don't concern the package directly and / or the answer depends a lot on the user's use case. For example:

• "What could I try if Pure CF outperforms Hybrid Recommendations?"
• "How could I incorporate time-based context information?"
• "What to do if only very popular items get recommended?"

Here, I would prefer to just link to the relevant Github issues within a pinned "Helpful Discussions" Issue. This way, we would acknowledge that the answer depends on the specific context and that strong trade-offs are involved. It would also be easier for users to propose other topics and useful issues via comments. (For now, I started collecting the issues in an RST document "Helpful Discussions"). What do you think @maciejkula?

[SimonCW]

Hey @maciejkula ,

Realistically, I currently don't have the time to go through more past issues. From my point of view this is a good starting point and ready to merge.

[maciejkula]

Thank you!