fix: Preserve offset for CFG edges when serializing to JSON

[cqc-alec]

Fixes #2592 . (Verified by regenerating JSON file using CLI built on this branch and rerunning test code.)
Fixes #2600 . (Verified by test included here.)

I don't think this counts as a breaking change to the JSON serialization format.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.07%. Comparing base (70ee1cb) to head (8796750).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2606   +/-   ##
=======================================
  Coverage   83.07%   83.07%           
=======================================
  Files         255      255           
  Lines       48422    48436   +14     
  Branches    43932    43946   +14     
=======================================
+ Hits        40225    40240   +15     
  Misses       6103     6103           
+ Partials     2094     2093    -1     

Flag	Coverage Δ
python	91.55% <ø> (-0.05%)	⬇️
rust	82.21% <100.00%> (+0.01%)	⬆️

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.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[aborgna-q]

This is too restraining.
The bug happens any time we have more than one non-df ports, so it would be safer to do

Suggested change

	let is_cfg_edge = op.is_dataflow_block();
	let offset = (is_value_port || is_static_input || is_cfg_edge).then_some(offset as u32);
	let multiple_other_ports = op.non_df_port_count(dir) > 1;
	let offset = (is_value_port || is_static_input || multiple_other_ports).then_some(offset as u32);

Alternatively, always encode the offset 🤷

It's not a breaking serialization change either way since the decoding logic works fine if the offset is already defined

[cqc-alec]

Yeah, my first idea was to encode the offset always, but this broke some round-trip tests in hugr-py for hugrs containing order edges.

I will try your suggestion.

[aborgna-q]

Ah, seems like hugr-py uses None only for order edges. (The fn comment seems wrong here, only order edges have offset -1)

hugr/hugr-py/src/hugr/hugr/base.py

Lines 971 to 980 in 70ee1cb

	def _constrain_offset(self, p: P) -> PortOffset | None:
	"""Constrain an offset to be a valid encoded port offset.
	Order edges and control flow edges should be encoded without an offset.
	"""
	if p.offset < 0:
	assert p.offset == -1, "Only order edges are allowed with offset < 0"
	return None
	else:
	return p.offset

If we wanted to follow that here then we can check

let other_port_is_not_order = op.other_port_kind(dir) != EdgeKind::StateOrder;
let offset = (is_value_port || is_static_input || other_port_is_not_order).then_some(offset as u32);

[cqc-alec]

Right, thanks. I've gone with your second suggestion. I tried the first, but ran into some round-tripping failures using testfiles exported by guppylang. With this second idea, all seems good. (Also checked that it still fixes #2592 .)