lattice(lat; kw...)

[pablosanjose]

As mentioned in the cancelled #93 we need to complete the ways we provide to change a lattice, or to build one lattice from another. At the moment we cannot change the Bravais matrix, the type, the embedding dimension or the sublattice names of a lattice after creation. This PR, that builds on #94, does just that. It provides a method lattice(lat; kw...) where kw are any of the kwargs for lattice creation (bravais, dim, type and names). To do this elegantly we needed to move bravais to a kwarg, which was done in #94.

From the implementation point of view, the key of this PR is to keep the mutation as efficient as possible. If we're only changing Bravais vectors, there is no need to touch the site position array or sublattice offsets. In contrast, if we're changing the dimension or the position type we need to allocate a whole new Unitcell. This done via parametric dispatch.

An open question here is whether we want also a hamiltonian(ham; kw...) or not. The keywords of hamiltonian are orbitals and type. Changing any of the two (unless only change the orbital names, but not their number, which is kind of a trivial change) implies recomputing the Hamiltonian (i.e. reapplying the model to the lattice, which often does not make sense without changing the model itself), so I'm inclined not to add such a method.

[codecov-commenter]

Codecov Report

Merging #95 into master will increase coverage by 0.20%.
The diff coverage is 84.00%.

@@            Coverage Diff             @@
##           master      #95      +/-   ##
==========================================
+ Coverage   62.15%   62.36%   +0.20%     
==========================================
  Files          16       16              
  Lines        2603     2617      +14     
==========================================
+ Hits         1618     1632      +14     
  Misses        985      985              

Impacted Files	Coverage Δ
src/convert.jl	53.84% <20.00%> (+3.84%)	⬆️
src/hamiltonian.jl	74.74% <100.00%> (ø)
src/lattice.jl	81.42% <100.00%> (+1.13%)	⬆️

---

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 37b2dba...3593331. Read the comment docs.

[pablosanjose]

Another future evolution of this is to finally implement Base.push!(lattice, site), which we currently don't have. Since that doesn't change the lattice type, it should be a simple matter of splicing the site (or sites) and recomputing the sublat offsets. And then, we can also consider merge!(lattice, sublattices)