Add average expression of neigbour markers

[ellispatrick]

We should add the average expression of the 5 nearest neighbours (that aren't that cell type) to the regression models. We should do this in addition to weighting for contaminated cells.