fix(ci): repair op-bench after graph_random_walk signature change

[Levalicious]

The benchmark job went red on #244. Root cause: avoidCycles (#243) added an argument to graph_random_walk, but native/op_bench.c still used the old 8-arg call — the runner's merge-ref (new store sources + old harness) failed to compile. Reproduced locally; the pre-change tree builds and benches fine.

• native/op_bench.c: pass the new avoid_cycles arg (0) — the harness keeps measuring the same op, so base/head stay comparable

Verified locally: make -C native bench compiles, runs, emits the JSON report (exit 0). On this PR the base-build step self-skips (base lacks the harness fix); the full base-vs-head comparison resumes on the next PR.

Follow-up note: the head-build step currently sends make output to /dev/null, which made the red run undebuggable — a one-step fix to print the build-log tail on failure is prepared locally but held back: pushing workflow files requires a token with workflow scope.

🤖 Generated with Claude Code

[github-actions]

Per-op benchmark — p50 cycles (base vs head)

base runs: 5, head runs: 5 (median of mins) · freq-correction ×0.9987 · threshold head−base > max(40 cyc, 5%)

op	base	head*	Δ	Δ%
lookup	124.3	125	+0.7	+0.6%
read_entity	40	40	+0.0	+0.0%
neighbors_d1	791	796	+5.0	+0.6%
neighbors_d2	2584.8	2598	+13.2	+0.5%
find_path_d6	180737	184056	+3319.0	+1.8%
search	1180214	1174832	-5382.0	-0.5%
entities_by_type	14038	14020	-18.0	-0.1%
random_walk_d5	1339.3	1342	+2.7	+0.2%
inc_walker_visit	17.5	17	-0.5	-2.9%
structural_sample	2595824	2565505	-30319.0	-1.2%
compute_merw_psi	928599	914506	-14093.0	-1.5%
create_entity	858	857	-1.0	-0.1%
create_relation	784	783	-1.0	-0.1%

* head frequency-corrected to base via the reference op.

No regressions beyond threshold.