Generalize unitcell and supercell to use generic SiteSelector machinery

[pablosanjose]

This was one tricky refactor, as it touches some of the oldest parts of Quantica. It generalizes the supercell routines to take arbitrary SiteSelectors. Before, supercell just took a region kwarg that was used to constrain any non-periodic direction of a superlattice (this then extended to unitcell too). With this PR we can specify other constraints, such as indices or sublats when building a new supercell or unitcell.

As an example application, we can combine this with the not wrapper from #97 to create vacancies by index (i.e. instead of by position). First create a Hamiltonian without vacancies, and plot it to locate the indices for the desired vacancies

using VegaLite
h = LatticePresets.honeycomb() |> unitcell(region = RegionPresets.circle(10)) |> hamiltonian(hopping(1))
vlplot(h)

Now, just rebuild the unitcell of h excluding the desired site with indices

h´ = unitcell(h, indices = not(549))
vlplot(h´)

[codecov-commenter]

Codecov Report

Merging #98 into master will decrease coverage by 0.05%.
The diff coverage is 94.52%.

@@            Coverage Diff             @@
##           master      #98      +/-   ##
==========================================
- Coverage   62.07%   62.01%   -0.06%     
==========================================
  Files          16       16              
  Lines        2634     2651      +17     
==========================================
+ Hits         1635     1644       +9     
- Misses        999     1007       +8     

Impacted Files	Coverage Δ
src/tools.jl	56.33% <66.66%> (-2.08%)	⬇️
src/model.jl	72.54% <88.88%> (+0.12%)	⬆️
src/lattice.jl	81.53% <96.49%> (+0.10%)	⬆️
src/hamiltonian.jl	74.82% <100.00%> (+0.08%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2587888...e5a3c8f. Read the comment docs.

[pablosanjose]

This includes a redesign of the supercell internals (the old ribbonfunc stuff), which (1) was rather obscure and (2) could lead to inconsistent/unexpected results, such as unitcell(lat) resulting in a unit cell different from lat.unitcell (!!).