Improving magnetic field memory access

[jgalan]

The bottleneck in ray-tracing calculations is coming from the integration of the axion-field inside an inhomogeneous magnetic field.

Mainly, the reason why the computation is expensive is due to the evaluation of the magnetic field in the class TRestAxionMagneticField, the required computation time may be evaluated using the macro REST_Axion_FieldIntegrationTests.

One could also test by accessing N-times to the TRestAxionMagneticField::GetMagneticField at random positions.

A failed attempt to improve the memory access efficiency is in the following PR. Where I replaced the nested std::vector scheme by a single std::vector.

#83