Compare performance of storing all link fields in one file vs one field per file

[konard]

[konard]

Looks like it will be similar to parallel array from https://en.wikipedia.org/wiki/Data-oriented_design

[konard]

We have SplitMemoryLinks in contrast with UnitedMemoryLinks where Data is split from indexes. We can also split indexes file into two files (one per index). Furthermore, we also can test parallel array implementation where even data itself is split into two files (one for sources, and one for targets). As well as, each index file can be split into one file per field.