Clarify mesh conventions

[adematti]

pmesh paint() and readout() seem to project particles in [0, boxsize] to [0, nmesh].
Yet, as defined in pmesh, x-coordinates larger than nmesh//2 are flipped to [- nmesh//2, 0].
Hence this function: https://github.com/adematti/mockfactory/blob/7df2a96bee83db561a64188a2c5491777852c2ce/mockfactory/gaussian_mock.py#L49
To keep current pmesh x-coordinates, one could also paint positions in [-boxsize/2, boxsize/2] (then the [-boxsize/2,0] part is shifted to [boxsize/2, boxsize] in the mesh array), e.g. bccp/nbodykit#659, but this is somewhat counter-intuitive (to me).
It'd be great to stick with one convention in the clustering pipeline.
@pmcdonal, any suggestion?

[pmcdonal]

This is the kind of thing... I'm not smart enough to stop right now and figure it out out of context... I'm sure I'd have an opinion if I got to using it, but I'd try not to freeze in any uncertain choice unnecessarily...