Let merge_datasets_as_delayed merge >2 datasets, filter by predicates and subset column

[mlondschien]

Description:

This Draft PR tries to implement changes as suggested in #235. The current philosophy is to

• Implement an additional function (e.g. merge_many_datasets_as_delayed) not to break the existing API
• rewrite align_datasets to efficiently cope with multiple datasets
• Implement an additional merge_many_dataframes or similar to the MetaPartition class to allow for merging of dataframes.

The new merge_many_datasets_as_delayed can be supplied with predicates to subset rows and columns to subset columns. To account for the case where a column is defined in multiple datasets, columns needs to be a dictionary of type {uuid: table: [columns]}.

[mlondschien]

Closing this as its superseded by #243 .