Feature/writers improvement

This generally does not reduces the file sizes, but it reduces the
writing and reading times (no encoding/decoding is required)

Also functional in parallel

mostly gather, gatherVariable and allGatherVariable

These functions return the number of items (template parameter)

The aim of this variant is to provide a more classic version of the
output: one value per cell or per node.

It can only be used for 1d data.

It remains to make it work in parallel.

This is useful/necessary to define properly cuts