Writing result zip takes too long

[brodmo]

With larger submission sizes (>500) zipping and writing the output makes up the bulk of the runtime and takes overall way too long. The main reason for this is probably that a file is written for every comparison pair. Writing the result should be made faster or made skippable, ideally both.

[tsaglam]

With the current implementation, it takes ~7 seconds on my M1 MacBook with the board game dataset (434 submissions, with a mean size of 1529 LOC) and -n 10000. Overall runtime ~35 seconds (14 seconds parsing, 11 seconds comparison, 2 seconds clustering). Thus, I would consider it closed for now.