Fix issues from a merge between merge-5.2-staging and main

[ncik-roberts]

Fix some tiny issues observed in this part of the merge:

-- The following is in effect a single merge (see below):
bac195af1fb487ccea799943f5666403bd78bfae dynlink dune again
8691c427ef0aee9181a2ebe655ea2ce631b22f5c Add jkind_axis to modules_without_implementation
764f730e5da69c373a7c03977e01b56893e9062f Dubious merge fixes
3970ad5e7ca99cfb1e60dd8c5e256669fe1ff726 Merge fixes
4b06f1928848af7c31aabe74f26a3f5a9b8818a5 Fixes to Select_utils required by previous merge => mshinwell confirms this changeset is ok
b180783b26b342885ee267ccfb3fa14778263c2d Merge remote-tracking branch 'flambda-backend/main' into merge-5.2
  => merge should be redone without resolving conflicts and then diffed directly against
     bac195af1fb487ccea799943f5666403bd78bfae as not all fixes were committed with the merge itself (but some were)

described here.

Mostly:

• use Chris's implementation of unboxed tuple printing
• use consistent type-checking for Pexp_unboxed_tuple and Pexp_tuple
• allow local opens in unboxed tuple types
• add an (I think) dropped profile_counters_function.mli. I'm least confident about this one.

[ccasin]

LGTM