Switch to awkward arrays and boost histograms

[cerminar]

This is a major update of the whole stack.

• all underlying libraries have been updated to the latest greates
• Collections now load data in awkward-arrays (pandas data-frame are only used for printing)
• histograms are now filled using boost-histograms in their "hist" implementation
• selections are not stored as string but as functions
• main script is

The resulting boost in performance is significant: the code O(10) times faster.
Some bits might still need to be migrated but most of the code is functional.

NOTE: still not using arrays of records with "behaviors"....this is probably next in the pipeline of changes