fix(ssa): Avoid going through i128 when casting signed to u128

[aakoshh]

Description

Problem*

Resolves #9926

Summary*

Changes the codegen for casting signed numeric types to unsigned to avoid going through the sign extended signed type, and rather go directly to the unsigned type after sign extension.

Changes simplify_cast to recognise when we are casting back and forth and just use the value as-is.

Disallows i128 in SSA

Additional Context

Casting from e.g. i8 to u128 worked like this:

1. sign extend i8 to i128
2. cast from i128 to u128.

This isn't wrong, but the front end doesn't allow i128, and while FunctionBuilder::insert_cast calls validate_numeric_type which did allow it, the SSA interpreter didn't expect this type and returned an error.

Sign extension itself works by casting to unsigned and adding the offsets necessary to transform the value to the correct two complement based on the target bit size. In itself the sign-extended value is already unsigned, so what was really going on is

1. cast from i8 to u8
2. cast from u8 to u128
3. cast from u128 to i128
4. cast from i128 to u128

In this form it is clear that all we need to do to avoid i128 appearing in the SSA is to leave out the last 2 steps.

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 number of Brillig opcodes executed

Generated at commit: 6a6d6fda50641fd3ba2361419f57de208be779a0, compared to commit: e5acaf603193ddd5136251516c463ae6c50997b4

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
unary_operator_overloading_inliner_min	-15 ✅	-27.27%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
unary_operator_overloading_inliner_min	40 (-15)	-27.27%

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: 6a6d6fda50641fd3ba2361419f57de208be779a0, compared to commit: e5acaf603193ddd5136251516c463ae6c50997b4

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
unary_operator_overloading_inliner_min	-11 ✅	-18.03%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
unary_operator_overloading_inliner_min	50 (-11)	-18.03%

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Test Suite Duration'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: c861444	Previous: 3c29fd7	Ratio
test_report_zkpassport_noir-ecdsa_	2 s	1 s	2

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[vezenovm]

Looks like we could use a snap update and a master merge.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Execution Time'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 079f3b6	Previous: e5acaf6	Ratio
rollup-block-root-single-tx	0.003 s	0.002 s	1.50

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[aakoshh]

@vezenovm thanks, it's updated now.