Project new data into approximate tangent bundle

[maedoc]

in the readme, it is stated

For test_ind, you can either use integers, which will be interpreted as indices of nodes, or floats, which will be interpreted as positional encoding over the tangent bundle in the spectral domain

and I would have liked to evaluate the approximate vector field on a grid after estimated with data. Since I don't include the grid values in the data for rvgp, they are not part of the k-nn graph and thus the graph laplacian eigenspace.

is there a straightforward way to use the existing k-nn graph & pre-computed graph laplacian to encode new data to evaluate the GP?