Allow aggregating expression by multiple columns

[rfriedman22]

I extended hooke:::aggregated_expr_data() so that the user can specify multiple columns to aggregate by (e.g. both cell type and perturbation). I tested the updated function on default parameters and confirmed that it still returns the same result -- mean_expression, fraction_expressing, and specificity is equal for all values and the first column of the result is still cell_group if there is only one column. However, if multiple columns are specified, the result now puts all of those columns first, for example:

agg_expr <- hooke:::aggregated_expr_data(cds, c("perturbation", "cell_type_broad_abbrev")) %>%
  head()

returns:

  perturbation cell_type_broad_abbrev            gene_id gene_short_name
1     ctrl-inj                    RPC ENSDARG00000000001         slc35a5
2     ctrl-inj                    RPC ENSDARG00000000002          ccdc80
3     ctrl-inj                    RPC ENSDARG00000000018            nrf1
4     ctrl-inj                    RPC ENSDARG00000000019           ube2h
5     ctrl-inj                    RPC ENSDARG00000000068       slc9a3r1a
6     ctrl-inj                    RPC ENSDARG00000000069             dap
  fraction_expressing mean_expression specificity
1         0.002456332     0.002757203 0.051766341
2         0.002183406     0.001340389 0.075315424
3         0.086244541     0.083973716 0.077649102
4         0.021834061     0.019834487 0.016199531
5         0.003548035     0.002764963 0.008075803
6         0.036026201     0.034195728 0.027027247

[rfriedman22]

Per discussion with Maddy, the specificity scoring is now optional. This is the slowest step, and if we are aggregating by e.g. cell type and perturbation, we don't want specificity metrics to be stratified by perturbation status.