Resolving the phi and Ez near z = z_wall

[mrhardman]

We typically observe a diverging Ez = - d phi / dz as z -> z_wall. Analytical considerations lead us to expect phi ~ sqrt(z - z_wall) as z -> z_wall. Using a finite-element grid with uniformly spaced element boundaries is potentially a wasteful use of resolution when we require many points near the wall, and few in the centre of the domain.

An optimisation of the code which permits flexibility is to change the grid specification to allow non-uniformly spaced element boundaries.

Branch and PR where this is carried out will shortly be pushed.