Create snapshot mechanism

[leoyala]

A new class is needed to handle the computation of statistics defined in the snapshot schema. To simplify the design and avoid assigning too many responsibilities for one class. The computation of everything that is needed for one snapshot should be split across the following categories:

1. statistics: column stats, # rows & columns, duplicate fractions, etc.
2. metadata: creation date, schema hash, dataset name, dataset path, etc.
3. sketches: minhash, hyperlonglong, etc.

Each class should be able to compute everything that is needed based on the object created by any reader.