Add conduit parser

[awkure]

Sometimes when we want to parse large .xlsx files we stumble upon the problem of exponentially increasing memory consumption. This PR aims to deal with that problem by introducing conduit parser that attempts to run in constant memory.

Here're some benchmarks

And how they deal with memory

Default parser

Conduit parser

In large scale it does appear to run in constant memory, consider a file with 1 million rows

And here the comparison with python xlsx2csv streaming library on 1 million rows file (spolier, not much of the difference here)

$ time python3 ./bench.py --takeyourtime 
2988.90 real      2984.17 user         1.98 sys
$ time ./conduit --takeyourtime
2686.39 real       409.44 user       841.22 sys