Credit the chosen community and keep edge weights in the volume totals

[arpitjain099]

Two things in hypergraph_modularity.py.

Both _last_step variants pick a target community into best, set dct_A[v] = best, and then update VolA[m]. m is the loop variable, so the volume is credited to whichever candidate the search happened to end on. On a seeded 30 node hypergraph the returned partition changes once it is VolA[best].

VolA and Ctr are built with np.repeat(0, n), which is int64, and then accumulate HG.edges[e].weight. On a hypergraph with weights of 0.5 the volumes come out [0, 3] instead of [2.75, 5.5], because each += 0.5 truncates. np.zeros keeps them.

tests/algorithms is 41 passing.