bug in collecting labels

[Kizzzy]

indexing.py -> index_KvsAll

I think there is a bug here because the temporal knowledge graph is not the same as the static knowledge graph

1. In the static knowledge graph, it is correct to use (subject, relation) as the key to collect labels

2. However, in the temporal knowledge graph, the time element should be considered, because the input in the prediction task of a sample should be (subject, relation, time), that is, the keyword should be (subject, relation, time).

Using (subject, relation) as a key in a label collection will introduce some incorrect labels that should not be there, resulting in high predictive metrics in filter setting

[LeetJoe]

I also find this but in fact after I fixing this, the result almost the same if just considering the filter case.