refactor: compile Hugr using abstract graph datatype

[acl-cqc]

• Refactor parent outside HugrOp
• Make abstract Hugr/Graph type.
  • TODO: More renaming. e.g. the new Hugr could just be HugrGraph(edit: done)
  • Stores 2-way edges
  • separates parent from op, and separates out the root.
  • parent-ness is now set when the name is created, not when the op is set (which may be later).
  • Also contains Namespace
  • and may record 2 children of each op to order first in output. This is hackily used for Conditionals so
  • TODO: make that a list not a pair
  • TODO: since we now have setFirstChildren we can remove the freshNodeWithIO and newWithIO from HugrGraph and do that work in the compile monad.
• Hence, also drop ordering of HugrOp, remove index from Case.

What do you think of onHugr? My thoughts were

• (blah ->) HugrGraph -> (a, HugrGraph) becoming (blah ->) State HugrGraph a seems like a win
• but State HugrGraph () rather than HugrGraph -> HugrGraph seems a bit clunky in HugrGraph.hs
  • I could keep those as HG->HG if I added another function like onHugr (probably using modify \s -> s {hugr=f (hugr s)}), but is it worth having two onHugrs?

No actual splicing of Hugrs yet but everything we need for that is here.

A nice follow-up would probably be to move parent into the Compile monad, but nvm.

Done on main as that has the most test-cases for Hugr generation (compilation of many classical cases removed in stack-machine branch so would not be as well-tested there).

• TODO: failure in closures.brat (malformed Bratgraph?) needs turning into an issue