max_penalty: Compute the maximum penalty, possibly sparse
In antiphon/glbinc:

Compute the maximum penalty, possibly sparse

1
2
3

max_penalty(X, y, index, offset, add_intercept = TRUE, center, scale,
  alpha = 1, eps = 0.001, maxiter = 1000, verb = 0, use_glm = FALSE,
  ...)

`X`	design matrix without intercept
`y`	response 0/1
`index`	grouping
`offset`	offset
`add_intercept`	TRUE
`center`	the column means for standardization
`scale`	sd vector of columns for standardization.
`alpha`	lasso vs ridge weight
`use_glm`	use the glm-function?
`...`	passed on to c-methods