Modularize the dependency on statistics library

[harendra-kumar]

We should plan to migrate to using streamly for most pruposes. Streamly has faster routines for some tabular processing required in this package (e.g. sort, intersection, difference, split). Also, streamly has basic statistics functionality as well, streamly-statistics has more. Especially, once we have streamly-core, streamly will be a lightweight dependency depending only on boot libraries.

statistics is a heavy dependency and we do not need it often. We can add a flag so that the core package can be compiled without having to use statistics. In the core package all that we need is mean/median/stddev which we can get from streamly. For now, we can keep using these from statistics but modularize it in such a way so that statistics can be replaced. We can add a build flag to include more advanced statistics stuff that depends on the statistics library.