GreenFunction indexing refactor

[pablosanjose]

This is a major cleanup and refactor of the indexing machinery of GreenFunctions. It is relatively risky, with subtle behavioral changes.

User-facing changes

• We now consistently return an OrbitalSliceArray whenever the indexing a GreenSolution contains enough information to do so. That means "always" except when indexing with sites, as in g(0.2)[diagonal(sites(1))] or g(0.3)[sites(SA[1],2:3)]
• diagonal indexing now returns a Diagonal matrix or a OrbitalSparseMatrix
• the override keyword in densitymatrix and josephson is now called override_path
• Function siteindexdict is now exported (to be paired with orbaxes, which is exported and now also documented).

New features

• We can now index sparsely by using a hopselector using the sitepairs function. g(0.2)[sitepairs(range = 2)] will return a sparse OrbitalSliceMatrix with gᵢⱼ for any site j in cell zero and any site i within a distance of 2 to any i. We can also use sitepairs(::HopSelector). This functionality will allow efficient non-local mean field calculations in an upcoming PR. In the future, sparse observables such as current should move to using this mechanism too.

Internal changes

• SparseIndices: The sparse indexing machinery (both diagonal and sitepairs) relies on a common SparseIndices type, and a set of new codepaths for them. These are designed to be also used to index other objects o that just need to define getindex(o, ::CellOrbitals, ::CellOrbitals) methods. One example is EigenProduct, used by GS.Spectrum, which represents the product Ψf(ϵ)Ψ'*e^{iϕdn}' in terms of eigenpairs ϵ, Ψ without computing the product explicitly.
• Internalized GreenSlice output: A gs::GreenSlice is a lazy wrapper of a g::GreenFunction and some GreenIndices. Here we add a new field output, which is a preallocated AbstractArray that is the output of call!(gs, ω; params...). This is not really done for performance reason (the gain is small, as the main runtime is spent finding the GreenSolution g(ω; params...), but rather for organization purposes: the complicated issue of what kind of AbstractArray should be produced and of what size is now gathered cleanly in the GreenSlice constructor. Also it allows to introduce the getindex!(output, g::GreenSolution, inds...) method, which will be useful when we need to extend it in the general mean-field machinery (where g will be swapped for a DensityMatrix)

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

Attention: Patch coverage is 86.76845% with 52 lines in your changes missing coverage. Please review.

Project coverage is 92.59%. Comparing base (5a6b446) to head (f3e5213).

Files with missing lines	Patch %	Lines
src/specialmatrices.jl	59.57%	19 Missing ⚠️
src/tools.jl	76.92%	9 Missing ⚠️
src/types.jl	86.44%	8 Missing ⚠️
src/show.jl	30.00%	7 Missing ⚠️
src/observables.jl	86.11%	5 Missing ⚠️
src/slices.jl	93.10%	2 Missing ⚠️
src/greenfunction.jl	99.06%	1 Missing ⚠️
src/hamiltonian.jl	75.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #315      +/-   ##
==========================================
- Coverage   92.90%   92.59%   -0.31%     
==========================================
  Files          38       38              
  Lines        6678     6819     +141     
==========================================
+ Hits         6204     6314     +110     
- Misses        474      505      +31     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.