Squib::DataFrame: better xlsx and csv utilities

[andymeneely]

I think the "hash of arrays" idea works really well in Squib, but often I'm wondering if we could use more utility methods for this. For example, the data['Name'].size feels like a hack to me - it should just be data.rows or something.

Here's what I have in mind. I'll edit this list over time.

• Call it a Squib::DataFrame
• It acts like a hash of arrays. So it still supports the [] and returns arrays of values.
• It has an #nrows method that returns the maximum length of a column across all columns
• It has a #ncolumns method that returns the number of columns
• It has a #columns method that returns the column names
• It has a pretty_print method that works a lot like my to_json for better version control diff'ing. Can also print out a given card.
• It has its own trim_whitespace functionality so that doesn't get duplicated (still gets turned on or off by csv and xlsx).
• It has some metaprogramming that allows for dynamic methods to be declared based on snake-cased access to columns. For example data['Actor Name'] could be accessed by data.actor_name. We could even add another method like #bind_them! that dynamically declares methods in Squib::Deck based on column names. Or are these "too magical"?
• Provide other validation in its construction, like duplicate columns and stuff.
• Quantity explosion is implemented in this class so we're not duplicating code there either
• More custom functionality around auto-converting to integers or floats or whatever. Nobody's complained about this yet, but it could be nice for some narrow cases.