Modified the search strategy for when a point is not found inside and element. The new implementation uses the distrance from the nearest point as a measure of which element should be used in this case. In addiition I have modified the loc coordinate definition when it is not found to be in an element (within the tolerance) so that it lies within the [-1,1]^3 box thereby reducing oscillations of extrapolating the polynomial expansions outside this region. Note we are not currently ensuring it is withiin the local region of the prism or tet shape. Finally there was a bug with how the routine was storing the local coordinat which is now hopefully fixed