[GeoMechanicsApplication] Fix the return mapping in Mohr-Coulomb tension cutoff

[mnabideltares]

Description

There is inconsistency found in the return mapping for tension-cutoff return zone, and the tension-cutoff corner point zone.

The principal stress components sigma1 and sigma3 are mapped back, while sigma2 is kept constant. However, sigma2 needs to be mapped back for tension-cutoff.
This issue was found because the code gives different results than Plaxis.

1. Fix the return mapping of sigma2, for tension-cutoff
2. Add unit test(s)
3. Update the documentation