ToDo

#TODO

1) Report line number only for the debug build in the helpers functions.

2) Verify the possibility to not copy the SV from the GPU to the CPU when the ODE solver and the PDE solver are using the GPU.

3) Make the "vm_threashold" calculus dynamic inside the code. The idea is to calculate the value using only an input percentage (e.g: APD_90, APD_80)

4) Improve activation time calculus to use the maximum derivative (check Lucas's old function from 2018)
    - Optimize the activation time calculus
    - Parallelize with #pragma and avoid race conditions inside the "persistent_data" structure
    - Use simple arrays ? Will consume more memory, but it will be faster.

5) Remove depracated code from src/linear_system_solver_library/gpu_solvers_cublas_11.c. Use https://github.com/NVIDIA/CUDALibrarySamples/blob/master/cuSPARSE/spsv_csr/spsv_csr_example.c as reference. Change cusparseScsrsv2 to cusparseSpSV.

6) Find a way to write the INa, Ito and ICaL currents in the ToRORd and Trovato models.

#KNOW ISSUES:
The logger symbols are only exported to an executable if an static library linked to a shared library uses then. For now this is ok. But I think it will be a pain in future releases.
The GPU linear system solver is not working for purkinje-only simulations
When the minimum number of PMJs is not reached the solver will be in an infinite loop