Hardcoded BUFF_LEN with larger numbers of samples

[russcd]

Hi All.

I've been prototyping a pipeline with ngsld and discovered an issue. The hardcoded buff_len parameter in gen_func.hpp will cause the program to fail to correctly read files with large numbers of samples (n=6449 and using beagle-formatted GLs). I expect this is because a single line exceeds the buffer size.

When I increased the BUFF_LEN to 500000 and recompile, the program runs correctly. Of course, I don't think that's the solution you would want to implement, but hopefully this is helpful in identifying a potential problem. Happy to provide test files if helpful.

Cheers,

Russ