PBC multigrid optimization

[sunqm]

• Improve memory usage of the multigrid implementation.
• Improve performance for
  • density evaluation
  • XC matrix integration
  • nuclear gradients computation
• Add analytical computation of strain derivatives, replacing the finite difference routine.
• Refactor gradients and stress tensor code
  • the local part of pseudo potential are evaluated along with XC integration
• Update SCF iteration driver, constructing an explicit dict for dump_chk function