Predefined Dictionaries
Would a set of zstd dictionaries defined after looking at the data we have and stored in the start of the experimental pack lead to better compression?
It seems that many files in a git repository may be similar, meaning much of dictionaries may be redundant.