high resource usage when parallelizing

[jjuod]

Parallel version seems to eat lots of RAM, at least at higher resolutions. For now, default setting changed to para=FALSE and parallelization is to be used at user's own risk. Would be good to add some max input size/resolution recommendations for parallel mode.

[william-denault]

Agreed I will check that, and may be that could be done by the user. Or we could propose mutilple paralelmisation, i.e apply parApply or normal

[jjuod]

In addition: some systems appear to have linear algebra libraries which already implement parallelization. This means that many CPUs are reserved, but the calculations are relatively short and the overhead is not worth it. Recommend RhpcBLASctl, to limit BLAS threads to 1 or 2, as in commit 8124d2c