chore: move unsigned overflow check from acir/brillig to ssa

[asterite]

Description

Problem

Resolves #8162

Summary

Additional Context

When inserting binary operations into the SSA we currently transform them a bit. For example:

1. if we have unchecked_mul between fields we turn that to mul because checked/unchecked doesn't make sense for fields
2. if we have unchecked_mul between bools we turn that into mul because it can never overflow

We did this to reduce the noise in the SSA, so that we don't get to see unchecked_mul between fields and wonder why it's unchecked if it doesn't make sense for it to be unchecked.

But... like that, and especially for point 2, it means that boolean multiplication will be checked unless we don't check it specifically for bools.

So... I wonder if instead unchecked_mul should be used whenever two values are being multiplied and they can never overflow. That means that mul between fields doesn't make sense anymore, and unchecked_mul is what makes sense. And the same applies for bools.

But... for now in this PR I just made those checked operations unchecked.

---

A next step, which I'll try as a follow-up PR, is to try #8111 again but this time memoizing the maximum bit size of values as we traverse the instructions (to avoid recursing multiple times, maybe that was the cause of the slowdown).

---

Yet another thought: unchecked: false is kind of hard to understand because it's a double negation. I wonder if renaming unchecked to checked and flipping all bools would make things a bit clearer.

Documentation

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] 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: 8e20ac596288b2f60228b1e6a48e484c3fdf4f6c, compared to commit: ae008d31e289ddb086b4c69e3fd287c5167fc4dc

🧾 Summary (10% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
hashmap	-8 ✅	-0.02%	-8 ✅	-0.01%
array_oob_regression_7965	-8 ✅	-20.51%	-8 ✅	-0.28%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
hashmap	33,469 (-8)	-0.02%	94,499 (-8)	-0.01%
array_oob_regression_7965	31 (-8)	-20.51%	2,843 (-8)	-0.28%

[asterite]

Cool, it went from a regression to an optimization! (not sure why, though, but maybe unchecked operations have more consequences along the way...)

[vezenovm]

LGTM, I have wanted to move more of this shared logic into the SSA so happy to see it!