Support for Manhattan distance (L1)

[elb3k]

As many usages of knn_vector search done in high dimensional space. From posts like curse of dimensionality suggests that L1 distance is better metric in high dimensional vectors. KNN vector type supports just L2 and Cosine similarity.

I am asking L1 feature request for knn.space_type

[vamshin]

Hi @Elbek-Khoshimjonov. Would leave this thread as a feature request. Any one reaching this thread for L1 feature please +1. This will help us prioritize the feature.