472 grb set does not execute in parallel

[anyzelman]

This MR contains the following changes:

• bugfix: an earlier MR moved the grb::set functions of the reference and reference_omp backends to reference/io.cpp, but did not update the parallelisation guards.
• code improvement: rely on config::OMP::localRange instead of parallel-for with a chunked static schedule. In several cases related to calls to grb::set and vector construction this leads to significant speedups.

This resolves internal issue 472 and contributes to internal issue 479.

Thanks to Alberto Scolari for contributing!