refactor(packages): Better handling of number ranges in bibliographies

Sort of a hack for now, see in-code comments.
The whole global picture would be harder, so let's go for a quick win
although imperfect.

csl/core/engine.lua

@@ -22,6 +22,7 @@ local CslLocale = require("csl.core.locale").CslLocale
 
 local superfolding = require("csl.core.utils.superfolding")
 local endash = luautf8.char(0x2013)
+local emdash = luautf8.char(0x2014)
 
 local CslEngine = pl.class()
 
@@ -1229,6 +1230,24 @@ function CslEngine:_postrender (text)
       -- or a question mark. But it's ugly.
       text = luautf8.gsub(text, "([…!?%.]" .. rdquote .. ")%.", "%1")
    end
+   -- HACK: Numbers ranges.
+   -- This is a shortcut, deviating from the CSL specification.
+   -- One one hand, it uses the page-range-delimiter to format all number ranges:
+   -- CSL says it should be used for page ranges only, and that an endash is implied otherwise.
+   -- But honestly, this is against most good typographic practices (in French, at least,
+   -- the single dash is recommended, but consistency is key anyhow, so chapter ranges
+   -- for instance should use the same delimiter as page ranges).
+   -- On the other hand, doing such things that late in the process is not great.
+   -- Looking at citeproc-lua, it does a lots of things in cs:number processing,
+   -- e.g. "1 & 2" is valid, and in "ordinal" form, it should be "1st & 2nd".
+   -- We aren't implementing that yet.
+   -- I'm not sure (at a glance) that those other citeproc implementations handle
+   -- roman page numbers (which may occur in some book front matters), etc.
+   -- Needless to say, centuries in titles, etc. are another can of worms.
+   local range = self.punctuation.page_range_delimiter  or endash
+   range = range == "-" and "%-" or range -- escape hyphen
+   local dashes = "%-" .. endash .. emdash .. range
+   text = luautf8.gsub(text, "(%d+)[^%d]+%s*[" .. dashes .. "]+%s*(%d+)", "%1" .. range .. "%2")
    return text
 end