fix: Fix evaluation order of assignments with side-effects in their rhs

[jfecher]

Description

Problem*

Resolves #8262
Resolves #8337

Summary*

Changed evaluation order of assign-subexpressions to match the comptime interpreter.

Additional Context

@vezenovm mentioned here #8262 (comment) that we could also desugar the blocks to remove any internal statements beforehand. This would still be beneficial since we can still get confusing results by doing:

a.0[{ a = ([1,2], 3); 1  }] = 7;

ie. assigning within the lvalue will now assign after the [4, 5] assignment instead of before. This is probably unexpected though it should be internally consistent now at least. That being said, changing this to remove internal statements beforehand to remove all sub-expressions requires an entire compiler pass which would be a larger pass so I thought I'd get this out first.

Edit: We're still not completely consistent here. There are differences between the comptime interpreter and SSA/brillig in compiling the above example but both are incorrect.

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.

[vezenovm]

Sounds like we still have a failure for the case you posted in the description?

a.0[{ a = ([1,2], 3); 1  }] = [4, 5];

Can we add a failing test for this before merging with a link to a new issue? Happy to translate my comments on #8262 into an issue as well

[github-actions]

Changes to number of Brillig opcodes executed

Generated at commit: 401c92f07009b70135499cd9b77a395454cdaa68, compared to commit: f81925d6a4b7b216877da2c4db18f24bff196b27

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
regression_6674_3_inliner_max	+51 ❌	+4.49%
regression_6674_1_inliner_max	+12 ❌	+1.81%
regression_6674_2_inliner_max	+12 ❌	+1.81%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
regression_6674_3_inliner_max	1,188 (+51)	+4.49%
regression_6674_1_inliner_max	676 (+12)	+1.81%
regression_6674_2_inliner_max	676 (+12)	+1.81%
poseidon_bn254_hash_width_3_inliner_max	136,213 (+117)	+0.09%
regression_5252_inliner_max	751,177 (+600)	+0.08%
poseidonsponge_x5_254_inliner_max	151,129 (+120)	+0.08%
regression_5252_inliner_zero	914,673 (+600)	+0.07%
poseidonsponge_x5_254_inliner_zero	183,727 (+120)	+0.07%

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: 401c92f07009b70135499cd9b77a395454cdaa68, compared to commit: f81925d6a4b7b216877da2c4db18f24bff196b27

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
regression_6674_3_inliner_max	+51 ❌	+12.72%
regression_6674_1_inliner_max	+12 ❌	+5.83%
regression_6674_2_inliner_max	+12 ❌	+5.83%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
regression_6674_3_inliner_max	452 (+51)	+12.72%
regression_6674_1_inliner_max	218 (+12)	+5.83%
regression_6674_2_inliner_max	218 (+12)	+5.83%
poseidonsponge_x5_254_inliner_max	3,686 (+2)	+0.05%
regression_5252_inliner_max	3,973 (+2)	+0.05%
poseidon_bn254_hash_width_3_inliner_max	4,862 (+2)	+0.04%
poseidonsponge_x5_254_inliner_zero	2,912 (+1)	+0.03%
regression_5252_inliner_zero	3,246 (+1)	+0.03%

[jfecher]

@vezenovm added, bug is here: #8344. It's actually worse than I thought since we're not consistent between comptime & SSA. I'm considering it a separate issue though and it is rarer to trigger I think.

[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: 185ed45	Previous: b8d04d6	Ratio
test_report_noir-lang_noir_bigcurve_	289 s	233 s	1.24

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

CC: @TomAFrench