Unsorted configurations

[jagot]

When calculating matrix elements between two Slater determinants, the ordering of spin-orbitals in a configuration is important, since permutation of the orbitals amount to permutation of the columns of the determinant, which changes the overall sign of the matrix element. The suggestion is to remove the enforced canonical ordering of orbitals in a configuration, and adding a kwarg to the Configuration constructor that allows initial ordering the orbitals, but subsequently, the ordering is retained, e.g. when replace!ing orbitals, the substitutions are made in-place. This is well-defined for spin-configurations, but we still need to decide what happens when substituting from a multiply occupied orbital in a Configuration{Orbital}, i.e. what does replace!(c"1s2 2p", o"1s"=>o"2s") imply?

Related to #23 and #32