chore: implement is_zero in terms of AcirVars

[TomAFrench]

Description

Problem*

Resolves

Summary*

Throwing up an old PR I have, just want to check what CI makes of it.

Additional Context

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [Exceptional Case] Documentation to be submitted in a separate PR.

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: b2ce36cfd4f40d26dccf99a1cbc8d4f6c734bfd5, compared to commit: bc9a44f285e0569825a307b06ee8acd93461c87e

🧾 Summary (10% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
conditional_1	+1 ❌	+0.01%	0 ➖	0.00%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
conditional_1	6,917 (+1)	+0.01%	38,799 (0)	0.00%

[jfecher]

What is the status of this PR?

[TomAFrench]

I think it's ready for review. I'm not dead set on this going in but my feelings on acir_variable.rs vs generated_acir.rs is that we should encode the logic of various operations in acir_variable.rs as much as possible so that we can make use of the optimisations which are present there. (Granted we're not benefiting from it here)

In comparison generated_acir.rs is relatively "dumb" as it needs actual witnesses, etc. so we should avoid it where possible and just have it for handling black box function calls, etc.