Rewrite MPS types to use C instead of CR (and CL) and add ψ.center functionality.

[VictorVanthilt]

This PR changes the name of MPS bondtensors. The bond tensor for all MPS types is now called C and is equivalent to what was previously called CR.

A new property "center" of FiniteMPS is introduced. It can be accessed through ψ.center.

It will return a HalfInteger i:

• isinteger(i) → i is a whole number and indicates the location of the first AC tensor present in ψ.ACs
• ishalfodd(i) → i is a half-odd-integer (e.g. 3/2), meaning that there are no AC tensors, and indicating between which sites the bond tensor lives.
  For example: i = 3/2 (= 1.5) means that there lives a C between the AL on site 1 and the AR on site 2

The goal is to use this new center property to avoid unnecessary gaugefixing in methods like norm(::FiniteMPS) where before, the center gauge was brought to the first site, after which the norm was calculated. After this PR the norm is calculated using the (bond) tensor at ψ.center, without gaugefixing.

An important note is that ψ.C[center(ψ)] does not work, ψ.C should be indexed using integers and ψ.C[i] returns the bond tensor between sites i and i+1

[codecov]

Codecov Report

Attention: Patch coverage is 81.54506% with 43 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/states/finitemps.jl	69.23%	16 Missing ⚠️
src/states/infinitemps.jl	74.07%	7 Missing ⚠️
src/states/orthoview.jl	84.37%	5 Missing ⚠️
src/algorithms/ED.jl	0.00%	2 Missing ⚠️
src/algorithms/timestep/tdvp.jl	71.42%	2 Missing ⚠️
src/states/multilinemps.jl	50.00%	2 Missing ⚠️
src/states/ortho.jl	95.12%	2 Missing ⚠️
src/states/windowmps.jl	60.00%	2 Missing ⚠️
src/algorithms/derivatives.jl	50.00%	1 Missing ⚠️
src/algorithms/groundstate/vumps.jl	66.66%	1 Missing ⚠️
... and 3 more

Files with missing lines	Coverage Δ
src/MPSKit.jl	100.00% <ø> (ø)
src/algorithms/approximate/idmrg.jl	97.67% <100.00%> (ø)
src/algorithms/approximate/vomps.jl	72.72% <100.00%> (ø)
src/algorithms/changebonds/changebonds.jl	100.00% <100.00%> (ø)
src/algorithms/changebonds/svdcut.jl	92.72% <100.00%> (ø)
src/algorithms/changebonds/vumpssvd.jl	100.00% <100.00%> (ø)
src/algorithms/expval.jl	72.30% <100.00%> (ø)
src/algorithms/grassmann.jl	99.14% <100.00%> (-0.86%)	⬇️
src/algorithms/groundstate/gradient_grassmann.jl	71.42% <ø> (ø)
src/algorithms/groundstate/idmrg.jl	100.00% <100.00%> (ø)
... and 17 more

[Gertian]

In my opinion renaming CR to C might also be confusing since it's then unclear if C[site] corresponds to the C matrix left or right of that site.
(Admittedly the CR is also confusing but this was the original motivation for the R)

In hindsight it's probably most logical to allow indixation of the Cs with half integers. This way we get things like : state.C[1+1//2] which should be clear to anyone reading/writing the code.

[VictorVanthilt]

Nothing that good documentation (that is planned) can't fix. Its also in the docstring of FiniteMPS.
Storing CLs and only having the CR property, like it was before with FiniteMPS, was even more confusing imo.

[Gertian]

It surely was confusing. Essentially I looked into the code to see what it returned whenever relevant...
Anyways, good docs are indeed sufficient.

For the front end user half indexing might be very handy and user friendly though. If you and @lkdvos are open to this I can make a PR. Imo both can just coexist :)

[VictorVanthilt]

@Gertian indexing with HalfIntegers doesn't seem practical nor aesthetically pleasing to @lkdvos and me so we will decide against this. I updated the FiniteMPS docstring to explain the implementation of the center property. It also contains the convention of where C[i] lives (to the right of site i).