[demo][bug] Mismatch between item id and item features of neg_sample_df

[andafterall]

For example, t1.movie_id is negative sampled item id. But t2.genre is one feature which belongs to the related original positive sample:

Please verify the code: