Fix and generalize wrap

[pablosanjose]

The wrap function was seriously bugged. In particular doing wrap(h) would end up modifying h. Not good :-D

This PR fixes wrap, and generalizes it to be able to wrap along several axes simultaneously. So, a 3D Hamiltonian h can be wrapped along directions 1 and 3, with corresponding phase factors phases = (ϕ₁, ϕ₃), doing

wrap(h, (1,3); phases = (ϕ₁, ϕ₃))

[codecov-commenter]

Codecov Report

Merging #81 into master will decrease coverage by 0.06%.
The diff coverage is 27.39%.

@@            Coverage Diff             @@
##           master      #81      +/-   ##
==========================================
- Coverage   62.42%   62.35%   -0.07%     
==========================================
  Files          15       16       +1     
  Lines        2355     2614     +259     
==========================================
+ Hits         1470     1630     +160     
- Misses        885      984      +99     

Impacted Files	Coverage Δ
src/KPM.jl	65.82% <ø> (+2.55%)	⬆️
src/Quantica.jl	100.00% <ø> (ø)
src/diagonalizer.jl	53.84% <ø> (+1.09%)	⬆️
src/greens.jl	0.00% <0.00%> (ø)
src/lattice.jl	67.39% <ø> (+2.35%)	⬆️
src/tools.jl	67.87% <35.00%> (+4.47%)	⬆️
src/mesh.jl	93.63% <50.00%> (-0.29%)	⬇️
src/bandstructure.jl	90.84% <70.58%> (-1.47%)	⬇️
src/hamiltonian.jl	75.21% <79.16%> (+6.86%)	⬆️
src/model.jl	72.00% <84.00%> (+0.27%)	⬆️
... and 12 more

---

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 da23f39...9110e15. Read the comment docs.

[pablosanjose]

The PR uses in-place sparse broadcast of the form s .+= factor .* s´ for sparse s, s´. This is currently suboptimal. The ideal way, I believe, would be to use the sparse-specific function map!, but that is currently broken when aliasing source and target, see issue https://github.com/JuliaLang/julia/issues/37375