Refactor of LatticeSlice and Contacts

[pablosanjose]

This addresses the first part of the proposal in #232:

• Make LatticeSlice a general lattice slice object that can refer either to sites (SiteSlice), orbitals (OrbitalSlice), or orbitals grouped by sites (OrbitalSliceGrouped)
• Make CellIndices a general cell slice object that can also refer to either of the above
• Make LatticeSlice organize its different CellIndices with a Dictionary for efficient iteration and access
• Make SiteSlice, OrbitalSlice, OrbitalSliceGrouped, CellSites, CellOrbitals, etc simply aliases of the above objects
• Prepare the path towards OrbitalSliceArrays by basing SelfEnergys, Contacts, GreenFunctions etc on OrbitalSliceGrouped instead of SiteSlices.

The refactor has been far more tricky that expected, since it touches the internals of all Green solvers. The result, however, is a cleaner and conceptually clearer codebase, I think.

No new tests have been added for the moment, but some additional testing would be advisable.

[codecov-commenter]

Codecov Report

Attention: 37 lines in your changes are missing coverage. Please review.

Comparison is base (91af232) 92.40% compared to head (a7b3a18) 92.78%.

Files	Patch %	Lines
src/types.jl	88.69%	13 Missing ⚠️
src/slices.jl	90.32%	12 Missing ⚠️
src/greenfunction.jl	91.17%	6 Missing ⚠️
src/solvers/green/bands.jl	77.77%	4 Missing ⚠️
src/iterators.jl	88.88%	2 Missing ⚠️

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

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #234      +/-   ##
==========================================
+ Coverage   92.40%   92.78%   +0.37%     
==========================================
  Files          34       34              
  Lines        5477     5500      +23     
==========================================
+ Hits         5061     5103      +42     
+ Misses        416      397      -19     

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

[pablosanjose]

Mmm, so that was unexpected. The test failure on Linux was completely unrelated to the PR, but still worrying. It was produced by the FunctionWrapper inside AppliedEigenSolver. Said FunctionWrapper is a closure over a Hamiltonian which somehow was garbaged collected after exiting the apply function that creates the AppliedEigenSolver. This should absolutely never happen, so it seems FunctionWrappers is somehow not registering the closure root with Julia's GC properly. That's at least my current understanding.

The workaround is to call the wrapped eigensolver function once before exiting apply. This seems to solve the problem, but I definitely don't know why.

In any case, this is a completely unrelated issue, see #235

[pablosanjose]

The use of Dictionaries.jl required some methods that were missing, but they have now been added to that repo (andyferris/Dictionaries.jl#130), and a release has been tagged, so we no longer need the workaround.