Responses are blown up for simulations considering both 3D Velocity Model and topography when the damping type is Rayleigh damping

[Clifford-Yen]

When using Rayleigh damping, 1D soil profile works with or without topography, but 3D velocity model only works without topography. Otherwise, the responses would be blown up to astronomical numbers.

Interestingly, the mass damping, no damping, and BKT2 work for the same analysis.

This issue is still under investigation. Please avoid using both 3D Velocity Model and topography at the same time.

[Clifford-Yen]

After some testings, mass damping, no damping, and BKT2 not always work. In fact, although Rayleigh damping never work when considering both 3D Velocity Model and topography, it still generates different results (maybe only two types of result).

[Clifford-Yen]

The major problem of this issue should be fixed in 3b7754c. After fixing the incorrect memory access, the results generated with the same input file should be consistent now. However, the damping issue still persists. Lowering maximum_octant_level helps, but the worst performer is still the Rayleigh damping.

[Clifford-Yen]

#5 was created to separate the issue and to make it more clear.