fix: always use u32 for indexing arrays (2)

[asterite]

Description

Problem

Alternative to #10 to compare performance.

Summary

Additional Context

PR Checklist

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[github-actions]

Changes to circuit sizes

Generated at commit: f7a5e52f2c54a41ba239f833d85db4735e7b1ae5, compared to commit: 309404b501d935b7856e1788530cba44abe44ad2

🧾 Summary (10% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
test_sha384_256.json	+250,839 ❌	+654.10%	+488,439 ❌	+464.00%
test_sha512_256.json	+250,839 ❌	+653.82%	+488,439 ❌	+464.00%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
test_sha384_256.json	289,188 (+250,839)	+654.10%	593,706 (+488,439)	+464.00%
test_sha512_256.json	289,204 (+250,839)	+653.82%	593,706 (+488,439)	+464.00%
bench_digest.json	164,767 (+143,199)	+663.94%	343,076 (+279,138)	+436.58%
test_sha384_1.json	95,790 (+82,791)	+636.90%	205,131 (+161,679)	+372.09%
test_sha512_1.json	95,806 (+82,791)	+636.12%	205,131 (+161,679)	+372.09%
test_sha384_100.json	95,889 (+82,791)	+632.09%	205,255 (+161,679)	+371.03%
test_sha512_100.json	95,905 (+82,791)	+631.32%	205,255 (+161,679)	+371.03%

[asterite]

The other one is better.