compaso: fix sigman units, change sigmav units (sigmav3d unaffected)

[lgarrison]

sigman_com and sigman_L2com were having their values inappropriately multiplied by the box size in CompaSOHaloCatalog. These fields are the square roots of the eigenvalues of the second moment tensor formed from the unit vectors of the halo particle positions, so they should be dimensionless.

Also, sigmav{Min,Mid,Maj,rad,tan}_{L2com,com} were all being unpacked as redshift-space displacements (units of comoving Mpc/h). We unpack all other velocity fields as proper km/s, so let's do the same for those.

sigmav3d_{L2com,com} are unaffected, as are the eigenvectors.

[boryanah]

Lehman, super cool, but silly question (note I haven't tried running the new code)

Isn't halos['sigmav3d' + m['com']] defined lower and at the point of defining sigmavmid,maj,min we only have the raw version of that? Sorry if I am missing something obvious

[lgarrison]

The halo info loader is actually smart enough to know that if column halos['A'] needs halos['B'], then B needs to be loaded first. The _get_halo_fields_dependencies() function is responsible for that.

[boryanah]

Ah, yes, I vaguely remember now that this is how it worked. I think everything is in great shape.