MemoryError in calculate_embedding_shift

[jacobkimmel]

If the embedding attribute of a Velocyto object is set to a matrix with many dimensions, and many cells are present, this line in calculate_embedding_shift will try to allocate a huge array and fail with a cryptic MemoryError.

velocyto.py/velocyto/analysis.py

Line 1704 in 0963dd2

unitary_vectors = self.embedding.T[:, None, :] - self.embedding.T[:, :, None] # shape (2,ncells,ncells)

I'd suggest adding a check that the resulting EmbeddingDim, Cells, Cells array fits within a reasonable amount of memory, and printing a warning if it seems too large (i.e. any matrix that would require >32GB of memory could print a "WARNING - Embedding contains N dims and there are M cells, this is a huge matrix that may lead to a memory error").