InverseField failing for some cases

Hello all,

I have added a very simple test in branch [bugfix-inverse-maps](https://github.com/gridap/Gridap.jl/tree/bugfix-inverse-maps) that showcases a problem I've found when inverting cell_maps from models with linear geometries.

Without considering that the cell_maps should be `AffineMap`s instead of `LinearCombinationField`s (for which I've opened another issue), the inverse-map algorithm should probably be converging for such an easy case.

The failing testing case can be found [here](https://github.com/gridap/Gridap.jl/blob/fa3b109f7fb5697bc1e2cb40e0eefa60ae80ace9/test/GeometryTests/InverseCellMapsTests.jl).

Thanks!

Start with test/GeometryTests/InverseCellMapsTests.jl and reproduce the failing linear-geometry case from the bugfix-inverse-maps branch. Read the inverse cell-map algorithm and compare its handling of LinearCombinationFields with the issue's AffineMap context. Done means the supplied test converges and passes for the reported case.

Written by the indexing model from the issue text.