feat(perf): Remove unused last loads in mem2reg

[vezenovm]

Description

Problem*

Resolves

Summary*

Follow-up to #5865

Make the last loads removal smarter by checking whether that last loads is actually used anywhere. If it is not, remove the loads and allow stores to that loads address to be removed as well.

Additional Context

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]

After this PR:
Even after removing this copy optimization we get a Brillig bytecode size improvement from 55 -> 44.

Keeping the copy optimization we get the following final SSA which produces 24 brillig opcodes:

After Array Set Optimizations:
brillig fn main f0 {
  b0():
    jmp b1(u32 0)
  b1(v6: u32):
    v27 = eq v6, u32 0
    jmpif v27 then: b2, else: b3
  b2():
    v28 = add v6, u32 1
    jmp b1(v28)
  b3():
    return Field 2
}

[vezenovm]

Ok this PR now produces the optimal SSA specified above without the copy optimization. Just getting a failure on uhashmap that is the same as #5897.

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: 85faf641c1cba5d018c805e76b45c58875ffeb60, compared to commit: 0db5610b501013e8b549102815f909aa0b656db7

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
poseidon2	-15 ✅	-4.52%
brillig_loop_size_regression	-31 ✅	-56.36%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
u128	3,794 (-4)	-0.11%
hashmap	36,837 (-105)	-0.28%
uhashmap	24,210 (-97)	-0.40%
fold_complex_outputs	1,015 (-5)	-0.49%
fold_numeric_generic_poseidon	1,114 (-15)	-1.33%
no_predicates_numeric_generic_poseidon	1,114 (-15)	-1.33%
poseidon2	317 (-15)	-4.52%
brillig_loop_size_regression	24 (-31)	-56.36%

[vezenovm]

Closing in favor of #5925