Seven-bias carbon-chain TranSIESTA/TBtrans examples

Replot actual reference data: bash view.sh
New calculations on an available CPU partition: sbatch job.slurm
The CPU partition template follows the whole-node allocation policy; the
workflow itself uses at most8physical cores,4concurrent2-rank MPI cases.
Check available resources before submitting. CPU module compatibility does
not constitute a timing benchmark for a specific CPU node.

Sequence: lead C8 k32/k64 -> zero-bias C32 TranSIESTA -> two TBtrans meshes
-> copy zero-bias DM/TSDE -> six nonzero-bias selfconsistent TranSIESTA cases
-> two TBtrans energy meshes for each Hamiltonian -> sisl and independent
Fermi-function integration -> actual transmission/current CSV and figures.

PBE/DZP,1.27Angstrom fixed C-C distance,12Angstrom transverse cell,150Ry,
300K electronic smearing. Biases -0.20,-0.10,-0.05,0,0.05,0.10,0.20V.
Energy steps0.010/0.005eV,window[-1.5,1.5]eV; current is one spin block;
nonmagnetic total current doubles it. Do not double a general spin-polarized
current. This is a finite numerical example,not converged experimental carbyne.

Environment: SIESTA/TranSIESTA/TBtrans5.4.2; Intel oneAPI2024.2, IntelMPI2021.13,
MKL LP64 sequential, sisl0.16.4. C.psml is read from the existing PseudoDojo0.4
PBE SR public installation. The pseudopotential is not redistributed here.
The process-local Intel logical HSX reader affects this script only.

Each completed coarse device.TBT.nc is moved to device-coarse.TBT.nc before
running the fine mesh. TBtrans refuses to overwrite the old NC. Keep both
files and compare their currents; do not delete the result you still need.
