streaming

[timoxley]

Should be able to get constant memory usage if columnify processed the data as a stream rather than loading everything and processing it all in one go.

Plan is:

… analysing the first N lines (100?) to gauge appropriate widths then simply use cell wrapping from then on.
… for larger datasets all the widths are likely going to hit the maximum anyway so this is actually a non-problem [for column widths]

See npm/npm#6477 (comment)