Dynamically join Buckets in data drift clustering

[sourabhagr]

During data drift calculations, we assume a fixed number of buckets and split the data equally among them. This causes issues when the data can have very few unique values. Ex: if the variable is tagged as continuous and it only has values = 0 and 1 in reference dataset, we should ideally have 4 buckets - <0, 0, 1, >1. But the current logic will split the data in 20 buckets with many buckets = 0 and remaining as 1