Wrong scheduling in MoL when using adaptive_stepsize = yes

[V M]

The functions used in adaptive stepsizing are scheduled in the wrong place, leading to segmentation faults when trying to use one of the RK methods that support adaptive stepsizes (RK45, RK45CP, RK65, and RK87) with adaptive stepsizing.

This can be fixed by changing the scheduling from:

schedule MoL_InitAdaptiveError IN MoL_Evolution AFTER MoL_PostStep

to

schedule MoL_InitAdaptiveError IN MoL_Evolution AFTER MoL_Step

‌

[Roland Haas]

The TestMoL thorn tries to test adaptive stepsize methods but does not enable “MoL::adaptive_stepsize” which means the bug was not caught.

[Roland Haas]

Pull request is here: https://bitbucket.org/cactuscode/cactusnumerical/pull-requests/12/mol-move-initadaptiveerror-so-that-error/diff

[Roland Haas]

Please review.

[Roland Haas]

Applied as git hash 13fb5ad6 "MoL: move InitAdaptiveError so that error estimate can be computed" of cactusnumerical