Added missing dependency between recv/xv and pair/bh_density.

During the first (fake) step of a run, but also when a cell has no active hydro particles, recv/rho is not activated. A foreign cell then depends on recv/xv to receive the drifted particle positions and (old) densities. However, a missing direct dependency between recv/xv and pair/bh_density for foreign cells would occasionally cause the latter to be executed before or even during the recv/xv. In the former case, the gas particle in the pair interaction would be uninitialised during the first step and undrifted in later steps. Not sure what would happen in the second case. This merge request adds the missing dependency and fixes this issue.

Note that this issue does not affect stars pairs, because those also depend on the sort, which is correctly activated for cells with no active hydro.