closure-Debug: transitive has_fn_fields closure → algebra tower derives Clone-only (E0277 −322)

[briansrls]

closure-Debug: transitive has_fn_fields closure (Lane A, carrier-independent emit-tail)

has_fn (04_emit_info.dag build_type_summary) only flagged a direct arrow field,
so it caught Magma { op: fn } but missed the tower that embeds it —
Semigroup{magma} → Monoid{semigroup} → CommutativeMonoid → Group — each of which still
transitively carries dyn Fn and therefore cannot #[derive(Debug/PartialEq/serde)].

Fix: transitive closure close_fn_fields (DFS over field_type_map edges, cycle-guarded),
applied in build_emit_graph_info (04_infer.dag). After it, the whole algebra tower emits
#[derive(Clone)] only — verified by execution (805 files emitted, 0 diagnostics; all five
tower structs Magma/Semigroup/Monoid/CommutativeMonoid/Group now #[derive(Clone)]).
.dag authority + hand-mirrored Rust seed (no regen gate yet, #5514 deferred).

Scoreboard (HostNative seed census, by execution)

• E0277 (trait-not-satisfied on derive): 982 → 660 (−322) — the closure-tower derives that
  could never hold are no longer emitted.
• total cargo errors 15,642 → 15,353 on the same forced-HostNative tree.
• E0119/E0392 deltas are main-merge drift, not this change: removing a derive can only drop
  E0277 or surface an E0369 ==; it cannot create a conflicting-impl (E0119) or unused-param
  (E0392). (Clean attribution via a same-tree control bypassing the close_fn_fields call.)

Honestly re-surfaced: cross-representation == on the closure-tower (E0369)

Removing the (always-broken) PartialEq derive turns its == call-sites from
E0277-on-derive into E0369-on-==. These are pre-existing — the impl never held — now
visible at the call site. They belong to the value-grounding == class (same family as the
operator's Value::Null / numeric-tower cross-representation-equality thread), and need a value
owner (stern-bat-782 is archived; capturing here so they have a home).

35 sites, all on CommutativeSemiring<Magnitude>-bearing types, 9 files:

• std_measure.rs:127,139,149
• std_approximate_field.rs:36,37,40,41
• std_realization_measurement.rs:56,85,90,144
• std_realization_schedule.rs:38,39
• std_realization_width.rs:73,92,112,116
• extdeps_cpu_types.rs:59,60,62,67
• product_compute_fabric.rs:250,314,321,585,591,597,602,1444,1451,1466
• product_placement_supply.rs:21,22,23
• v2_workflow_ci_floor_plan.rs:371

(The v2_*/openapi == E0369 deltas in the raw census diff embed no closure and are
baseline drift, not attributable to this change.)

CI

fmt gate fix folded in: the hand-written close_fn_fields fold closure in the seed mirror was
not rustfmt-canonical (block-match → expr-match); logic byte-identical.

[gunbai-bot]

CI red is 100% inherited main breakage — not introduced by this PR

This PR changes exactly 4 files (04_emit_info.dag, 04_infer.dag, and their two Rust seed mirrors) — the transitive has_fn_fields closure. It touches none of the files the failing tests exercise. Verified by execution with a local cargo nextest run -p v1-compiler-tests (matching the CI gate exactly), then re-run of the 4 failures in isolation:

test	isolation	cause
parse::source_text_at_lookup_flat_in_file_size	FAIL (det., 0.011s)	#5567 (85a1e8b8f3, "Strip comments from .dag tree") made 02_parse.dag pure-ASCII → the test's assert!(!source.is_ascii()) non-ASCII-fixture precondition is now unsatisfiable
parse::tokenizer_non_ascii_performance_regression	FAIL (det., 0.012s)	same — #5567 ASCII-strip breaks the non-ASCII fixture precondition
source_audit::emit_handles_for_loop	FAIL (det., 0.010s)	#5575 (e2533bf58c, value-grounding Empty{} → Rc::new(vec![])) removed iter().cloned() from 05_emit_rust.dag, but the audit still asserts source.contains("iter().cloned()")
interp_recorded_fixture_test::clock_now_record_then_hermetic_replay_holds	PASS (36s)	load-induced Clock.Now freshness flake under the full-suite 47-slow-test contention (614s run); passes deterministically in isolation

All three deterministic failures reproduce on origin/main itself (confirmed via git show origin/main:…): origin/main's 02_parse.dag is ASCII and its 05_emit_rust.dag lacks iter().cloned() while the audit still asserts it. So the v1-compiler-tests nextest gate is red on main independent of this branch — a fleet-wide blocker that needs a dedicated test-drift fix (owners of #5567 / #5575), not a change here.

This PR's own change is green by execution: 761/765 pass, the algebra tower (Magma/Semigroup/Monoid/CommutativeMonoid/Group) now emits #[derive(Clone)] only, E0277 982→660 (−322) on the HostNative census. fmt (cargo fmt --all --check) and clippy (--all-targets -D warnings, whole workspace) both clean.