fix local env B related artificial spikes

[ahnyu]

This PR fixes a bug in the local environment calculation used by the assembly-bias parameters (Bcent, Bsat). The issue was that Menv was estimated independently within each superslab, so halos close to a superslab boundary were missing neighboring halos from adjacent superslabs. This led to incorrect Menv values near slab edges and introduced an artificial x-dependent striping pattern in the derived fenv field.
The fix is:

1. in prepare_sim.py, compute raw Menv for each superslab using padded halos from neighboring superslabs, save slim per-slab env files containing id, mass, and Menv.
2. in abacus_hod.py, build the final global fenv_rank from the concatenated env files and map it back to the staged halo and particle catalogs by halo ID.
   This fix is for box, the light-cone approach is unchanged.
   Included are before/after plots showing the clustering signal (Pk), galaxy counts along each axis.

• Before (note spikes around k = 0.11, 0.22, 0.33, slab boundaries):

• After:

[lgarrison]

Thanks @ahnyu! I think this sounds reasonable.

@SandyYuan Are you available to take a look at this?