[ENH] Generalized PSeAAC algorithm

[satvshr]

The current implementation of the PSeAAC algorithm was designed solely to satisfy AptaNet, utilizing a version of the PSeAAC algorithm that AptaNet had specified.

The purpose of this PR is to generalize this algorithm, as it is a very popular method in bioinformatics. The original paper has over 1500 citations, and a simple search on Google Scholar will show numerous papers referencing it in the context of aptamer design.

The problem with the current implementation is:

1. It is built to give best results when k-mer length is 4-which is incorrect. A k-mer of length 4 gives the best results with 18 physicochemical properties, not 21 as used in the current implementation.
2. It uses 21 properties for all k-mer lengths, which does not lead to optimal results, as each different k-mer length has a preferred number of physicochemical properties.
3. Grouping properties into groups of 3 is constraining, and frankly too rigid, for absolutely no reason.
4. There is no allowance for grouping arbitrary indices together; it currently only allows groups of 3 in ascending order.