Use generator instead of list in create_file_intersections() to save memory

[pwwang]

I have 70+ vcfs. While running consistency, it complains about memory usage with the list used in create_file_intersections(). Using a generator solves it.

[ACEnglish]

Thank you for the change. It all looks good.

The consistency tool could definitely use some further improvements. I've been meaning to refactor it for some time. If you find further opportunities to fix it, please feel free to create more pull requests.

[pwwang]

Will give it a try. Current implementation still requires tons of memory with large # vcfs.

[ACEnglish]

So, thinking about this, I feel like consistency tool could easily be sped up if we change one assumption. Currently, it has no requirements for input VCFs being sorted. If we require sorted/indexed VCFs, it would be crazy simple to reuse truvari.bench.file_zipper to simply 'pop' entries from multiple VCFs in-order and do the comparisons.

Given that your analysis of 70+ VCFs is likely the most use consistency has gotten, I'm curious about your opinion of requiring sorted/indexed vcfs. If you would be more than okay with that, I think I could prototype a much faster/lighter consistency over the weekend.

[pwwang]

I don't think this requires sorting/index of VCF files. See my implementation at #127