better pretty-printing of module shapes

[gasche]

(Remember: "shapes" are an abstraction of the module structure of OCaml programs introduced in #10718.)

I am trying again to understand what it is about the Irmin source code that made earlier implementations of shape-reduction blow up. This involves looking at the -dshape output for a manually-minized subset of irmin.ml, whose printing starts like this:

 "Maker_ext"[module] ->
     Abs<Irmin_short.167>
        (CA/1399,
         Abs
            (AW/1400,
             Abs
                (N/1479,
                 Abs
                    (CT/1523,
                     {
                      "Make"[module] ->
                          Abs<Irmin_short.166>
                             (M/1529,
                              Abs
                                 (C/1555,
                                  Abs
                                     (P/1574,
                                      Abs
                                         (B/1588,
                                          Abs
                                             (H/1589,
                                              {
                                               "AW"[module] ->
                                                   Abs<Irmin_short.111>
                                                      (K/1234,
                                                       Abs
                                                          (V/1235,
                                                           {
                                                            "clear"[value] ->
                                                                <Irmin_short.97>;
                                                            "close"[value] ->
                                                                <Irmin_short.95>;
                                                            "find"[value] ->
                                                                <Irmin_short.62>;

(I won't show the whole output because it takes 2GiB.)

This PR introduces a compact notation for n-ary functors (and reduces the structure indentation slightly). -dshape then gives the following result:

 "Maker_ext"[module] ->
   Abs<Irmin_short.167>
      (CA/1399, AW/1400, N/1479, CT/1523,
       {
        "Make"[module] ->
          Abs<Irmin_short.166>
             (M/1529, C/1555, P/1574, B/1588, H/1589,
              {
               "AW"[module] ->
                 Abs<Irmin_short.111>
                    (K/1234, V/1235,
                     {
                      "clear"[value] -> <Irmin_short.97>;
                      "close"[value] -> <Irmin_short.95>;

which is much nicer to read and work with.

[dra27]

LGTM - perhaps let @trefis or @voodoos have a chance to see prior to merging?

[voodoos]

Looks good to me too. It does read much easier than before.

[gasche]

Thanks! I must say that the existing pretty-printing code was rather good already. I rarely use Format and need some time to get back into it whenever, and there was an impressive display of various options used in the existing pretty-printer. I spent some amount of time trying to tweak the code in various ways, to generally worse results.