Rename normalize to subscale_by

[mjskay]

Per discussion here: #219 (comment)

I think something like subscale_by would be more descriptive and would be more consistent with the new subscale and subguide params.

• add subscale_by with options:
  • NULL or character() <=> normalize = "all"
  • "panels" (or maybe "panel") <=> normalize = "panels"
  • "xy" <=> normalize = "xy"
  • "groups" (or maybe "group") <=> normalize = "groups"
  • character vector = group by those columns in the layer data frame
• deprecate normalize
  • do the same thing I did with slab_type (keep an internal copy of the old param to detect use)
  • if someone uses normalize = "none" point them to subscale = "identity" instead
  • if someone uses normalize = "all" point them to subscale_by = NULL instead
  • otherwise point them to subscale_by = ...