Translation: resolve conflicts in remaining files (translmod, transclass, translobj, translprim, translattribute, switch)

[ncik-roberts]

Finish off what #179 and #171 started.

Notable resolutions:

• As explained in Translation: Resolve conflicts in matching.ml and translcore.ml #179, there's a new field on Pfield. An unresolved thread on that PR is also relevant here: Should accessing a module field be Reads_vary like it is upstream? I've asked Mark to weigh in.
• Lambda.lfunction was made a private type with a smart constructor both internally and upstream, and we have some extra arguments (region, mode). These are trivial to resolve.
• Two small diffs that we pull in from: Use Compilation_unit.t instead of Ident.t for globals flambda-backend#871
  • explains diff in translmod.ml where we don't need is_compilation_unit
  • explains diff in translprim.ml for add_used_primtive (where a change made upstream is irrelevant for us)
• Hardcoded checks for misplaced attribute warning go away internally, in preference for the more general framework in Chris's patch
• Threading through of kinds

Testing

for module in translclass translobj translmod translprim translattribute switch; do
  dune b --workspace=duneconf/boot.ws _build/default/.ocamlcommon.objs/native/$module.cmx
done

[goldfirere]

Hmmm.... maybe that was a bad idea because this was targeting a branch.

[ncik-roberts]

I think that's OK. There's only two small unresolved questions on the target that Mark will look at tomorrow, and I don't think this is blocking anything (it's just mls).