Wannier restart mode

[qiaojunfeng]

For the wannier90_qp step, 3 modes to restart

1. rewannierise with the GW-corrected eig file
   • need to sort amn/mmn/eig/... files by gw2wannier90.py
   • need to shift dis_froz_max since GW might shift CBM, otherwise difficult to compare with W90@DFT bands
   • no GW bands to compare band interpolation quality
2. sort chk file in additional to amn/mmn/eig/...
   • since GW only correct eigval, wavefunction unchanged, the wannier unitary matrices can be left unchanged
   • have more confidence in band interpolation quality
   • if chk is binary format, need to call w90chk2chk.x in gw2wannier90.py, not very robust
   • if amn/mmn/eig/chk files are large, would take a lot of time to run gw2wannier90.py
3. no sort, only add GW corrections to eig
   • the fastest, and robust
   • results should be the same as 2
   • I think wannier90 only need sorted eigval in the disentanglement step, in the Hamiltonian interpolation, we only Fourier transform the Hamiltonian matrices, no need to force them in order
   • the new eig file is unsorted, only add GW corrections, band index not changed

In workflow, since aiida<2.0 has no Enum type, I implement it as IntEnum so I can use orm.Int for spec.input
https://github.com/epfl-theos/aiida-yambo-wannier90/blob/4457c011fb264ca74c898df31e51ffbfcf3c239c/aiida_yambo_wannier90/common/types.py#L5-L20
https://github.com/epfl-theos/aiida-yambo-wannier90/blob/4457c011fb264ca74c898df31e51ffbfcf3c239c/aiida_yambo_wannier90/calculations/gw2wannier90.py#L62-L68
The current NO_SORT works fine, so I set it as default.