add function for testing nested hypotheses over an edgeR GLM fit

[russHyde]

For example,

nested_lrt <- function(fit, contrasts, contrast_tree, p_thresholds, ...)

[russHyde]

How to define the nested sets of contrasts?

# set_hierarchy
contrast_set | parent | p_threshold |
----------------------------
all | NULL
subset1 | all
subset2 | all
subset3 | subset1
subset4 | subset2

# set_contents
contrast_set | contrast
------------------------------
all            | c1
all            | c2
all            | c3
subset1        | c1
subset1        | c2
subset2        | c3
subset3        | c1
subset4        | c2

Or

L <- list(start_node, all_contrasts, subset1, subset2, subset3, subset4)
# where each of the entries is a list:
contrast_node <- list(parent, tree_level, contrasts, p_threshold, lrt, significant_features)

Or use data.tree - but I don't think it makes the process any easier than just iterating over a level-sorted list.