Closes #130

Introduces a new mincoordination kwarg in unitcell(::Hamiltonian,...). This removes sites until all remaining have at least mincoordination neighbors. The way this works is we iterate over the mask of the supercell to eliminate any sites that have too few neighbors. This pass is repeated until no more sites are removed. As the neighbor searching reused a large part of non-trivial hamiltonian unitcell code, I had to refactor the latter so that functionality could be shared.

Example from tests:

h = LP.honeycomb() |> hamiltonian(hopping(1)) |> unitcell(10, region = !RP.circle(2, (0,8)))
h´ = h |> unitcell(1, mincoordination = 2)
vlplot(h, size = 300, maxdiameter = 6)
vlplot(h´, size = 300, maxdiameter = 6)