Use a sparse(r) matrix

The politicians x followers matrix might be better compressed by using `Matrix::sparseMatrix` ([as in this script](https://github.com/pablobarbera/echo_chambers/blob/master/02_estimation/07-create-initial-matrix.r#L49)) instead of `Matrix::Matrix`. The current code is efficient enough, though.

Start by locating the politicians-by-followers matrix construction and compare it with 02_estimation/07-create-initial-matrix.r, especially the sparseMatrix example at line 49. Check that downstream operations accept the sparse representation, then verify that the matrix is built correctly and remains efficient.

Written by the indexing model from the issue text.