feat: Add CircuitHistory

[lmondada]

I'm putting this out here for discussion. My main question: how should I expose the hugr-like interface of CircuitHistory?. I've proposed an implementation of HugrView on a wrapper type ExperimentalHugrWrapper of the circuit history. See src/diff/experimental.rs.

A CircuitHistory is always equivalent to a valid Hugr (see CircuitHistory::extract_hugr), but implementing HugrView is difficult because

• there is no "base hugr", so implementing HugrInternals is impossible. This could be implemented if the base hugr was node specific, i.e. fn base_hugr(&self, node: Node) -> &Hugr
• the hugr::Node type is fixed, but in a CircuitHistory the "natural" node index is a pair (Hugr, Node), i.e. an index/ref to the correct hugr in the history, along with a node within it. Right now, I have "solved" this with the world's greatest hack, using the integer within Node to encode both a hugr index and the node index. This will overflow at 2^16 which we are bound to reach pretty quickly
• Enumerating all nodes and counting them requires traversing the hugr, so this is a slow operation. Same with edge_count

---

There are still some minor TODOs, before this is mergeable (if we wish to merge it)

• finish writing the history::test_history test
• generalise apply_rewrite to CircuitHistory

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 82.51%. Comparing base (6bf8e4d) to head (8478612).

Additional details and impacted files

@@                Coverage Diff                @@
##           feat/circdiff     #778      +/-   ##
=================================================
+ Coverage          82.48%   82.51%   +0.02%     
=================================================
  Files                 65       65              
  Lines               8178     8160      -18     
  Branches            7922     7904      -18     
=================================================
- Hits                6746     6733      -13     
  Misses              1012     1012              
+ Partials             420      415       -5     

Flag	Coverage Δ
python	82.42% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ss2165]

This is really cool.

I haven't looked at the code in detail but I will comment on the main question.

I think it is worth having a read only and writable traits on hugr-like structs. This history object is one such struct. Our best candidate for that trait is currently HugrView. I think it falls short of that requirement by depending on HugrInternals, which is a trait which only makes sense when backed by one Hugr.

Ideally I think we should break that dependancy, but I'm not sure how viable that breaking change is. @aborgna-q may be better placed to comment on that.