Optimize CsvInsight w/ striped reading/splitting

[mpenkov]

Currently, the preprocessor splits the input file into multiple parts (using split). This part runs on a single core, because the splitting in its current form cannot be parallelized.

Modify the splitter to run on multiple cores:

• Open N files, where N is the number of cores
• Start N subprocesses to read from the input file
• Each subprocess reads the input file entirely
• nth subprocess only writes lines where line_number % N == N