Second part of time-evolution methods for MPS

[claudius-hubig]

Hi Miles,

these commit contain the rest of the review part of our review – descriptions of the local Krylov/time-step targeting, the TDVP, the MPO and the global Krylov methods as well as the collection of tricks.

I've also made two changes to generate.jl which you may also wish to have in your new generator:

• indented lists in TOCs were not closed correctly, which lead to the "References" being at the same level as the last non-reference item instead of being a top-level item
• as generating the HTML takes longer and longer, I've added a (rough) check to skip unchanged files. Of course this should in theory also check the modification date of generate.jl itself as well as those of other supporting files, but in practice it works much faster and deleting the output folder is (at least for me) also an option if I want to make sure to regenerate everything.

The content isn't quite where it could be, I may get around to this in January but can't make any promises. Specifically, it would be nice to:

• add some parts of our MPS Sec. 2 to the generic MPS algorithms, in particular the variational orthogonalisation may be interesting to other users, too.
• improve interlinking with existing pages as well as within the newly-added ones: the web format clearly makes this easier, but the current generator a bit harder. It would be nice if we could label section titles to enable linking to anchors (i.e., rename anchors from #toc_1,2,3 to #errors and provide some way to link to mps/algorithms/timeevo/tebd#errors).
• add the experiments we did at the end: those especially can and should be extended by others if they come up with either another problem or another method

In any case, thanks again for your work on this project!

Claudius

[emstoudenmire]

This is all so great - thanks Claudius. I’ll go ahead and pull and if small fixes need to be made, make them after.