Refactor InfiniteEnvironments and GradientGrassmann

[lkdvos]

This PR consists of a number of changes that ultimately boil down to removing the automated recomputation of the infinite environments.

In order to achieve this, a number of additional changes are also included:

• most notably, I finally got around to cleaning up the grassmann implementation, which now uses the preconditioner in OptimKit. This GREATLY improves the clarity, and should hopefully be a nice change for future development.
• all infinite environments are now unified
• various utility changes
• various preparatory changes for having above and below different.

[codecov]

Codecov Report

Attention: Patch coverage is 91.04730% with 53 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/algorithms/grassmann.jl	81.41%	21 Missing ⚠️
...c/algorithms/excitation/quasiparticleexcitation.jl	45.45%	12 Missing ⚠️
src/algorithms/derivatives.jl	76.31%	9 Missing ⚠️
src/environments/multiline_envs.jl	85.36%	6 Missing ⚠️
src/algorithms/approximate/vomps.jl	80.00%	2 Missing ⚠️
src/environments/multiple_envs.jl	93.33%	1 Missing ⚠️
src/states/finitemps.jl	50.00%	1 Missing ⚠️
src/states/infinitemps.jl	87.50%	1 Missing ⚠️

Files with missing lines	Coverage Δ
src/MPSKit.jl	100.00% <ø> (ø)
src/algorithms/approximate/approximate.jl	100.00% <100.00%> (ø)
src/algorithms/approximate/fvomps.jl	95.65% <100.00%> (-0.58%)	⬇️
src/algorithms/approximate/idmrg.jl	97.77% <100.00%> (+0.10%)	⬆️
src/algorithms/changebonds/changebonds.jl	100.00% <100.00%> (ø)
src/algorithms/changebonds/optimalexpand.jl	100.00% <100.00%> (ø)
src/algorithms/changebonds/vumpssvd.jl	100.00% <100.00%> (ø)
src/algorithms/expval.jl	74.60% <100.00%> (+2.29%)	⬆️
src/algorithms/groundstate/dmrg.jl	100.00% <100.00%> (ø)
src/algorithms/groundstate/gradient_grassmann.jl	75.00% <100.00%> (+1.66%)	⬆️
... and 25 more

... and 2 files with indirect coverage changes

[lkdvos]

While debugging the approximate methods, I would say that we really need to have a proper look at choosing a convention for above and below, since now this is somehow different for the approximate methods compared to the leading_boundary methods... This also ties together with a proper debug of the whole Multiline machinery

[lkdvos]

LTS test failures might be unrelated: retrying CI on master to verify that here
EDIT: issue seems to have solved itself