feat(decompile): retain canonical SSA sidecar

[Jon-Becker]

What changed? Why?

Builds and retains contextual stack SSA inside the decompiler's canonical sidecar.

CanonicalAnalysis now keeps the ID-consistent tuple of:

• canonical Program;
• contextual abstract CFG and diagnostics;
• expression and state-version arenas owned by that CFG;
• contextual SSA blocks, phi nodes, and lowered effect operands; and
• selector-to-canonical-block mappings.

This makes canonical SSA available to upcoming function recovery and graph-based typed lowering without recomputing analysis or flattening calling contexts. Legacy source and ABI generation remain unchanged.

This PR is stacked on #751.

Notes to reviewers

• SSA is derived immediately after contextual analysis so all ExprId, StateVersionId, ContextualPoint, and SsaValueId relationships remain scoped to one CanonicalAnalysis artifact.
• This is still a sidecar migration stage. It intentionally does not switch production source lowering.
• The additional SSA construction cost measured in feat(vm): build contextual stack SSA #751 is 1ms for USDT/Uniswap V2, 8ms for Uniswap V3 SwapRouter, and 103ms for Seaport.
• Consumers should use points_for_selector and then index the corresponding contextual SSA blocks rather than projecting contexts onto a single block prematurely.

How has it been tested?

• cargo test -p heimdall-decompiler --lib — 75 passed
• cargo test -p heimdall-vm --lib — 204 passed
• cargo check --workspace --all-targets — passed
• cargo +nightly fmt --all -- --check — passed
• git diff --check — passed
• Extended the canonical-sidecar test to assert one SSA block per contextual entry state and verify that retained storage effects are available through the sidecar.

[github-actions]

❌ AI Evaluation Suite for 6ee877b

📊 View the full HTML evaluation report.
Completed heimdall-eval. View workflow run.

📊 View evaluation scores

Test Case	CFG	Decompilation
NestedLoop	100	10
SimpleLoop	100	15
TransientStorage	100	10
NestedMappings	100	48
SimpleStorage	100	96
WhileLoop	100	10
NestedMapping	100	2
Events	100	62
Mapping	100	96
SwapRouter	100	20
WETH9	100	82
UniswapV2Pair	96	12
TetherToken	98	32
Seaport	98	5
Average	99	35

• Run AI Evaluation Suite

[github-actions]

✅ Coverage Report for d7fe889

Metric	Value
Base branch	71.08%
PR branch	75.18%
Diff	+4.10%

[github-actions]

Benchmark for d7fe889

Click to view benchmark

Test	Base	PR	%
heimdall_cfg/complex	13.7±0.15ms	199.8±1.17ms	+1358.39%
heimdall_cfg/simple	1405.5±44.19µs	632.5±8.37µs	-55.00%
heimdall_decoder/seaport	58.4±2.91µs	56.3±1.99µs	-3.60%
heimdall_decoder/transfer	4.2±0.33µs	4.2±0.30µs	0.00%
heimdall_decoder/uniswap	16.0±0.74µs	15.8±0.55µs	-1.25%
heimdall_decompiler/abi_complex	57.0±1.12ms	277.5±4.06ms	+386.84%
heimdall_decompiler/abi_simple	1497.2±32.20µs	2.1±0.01ms	+40.26%
heimdall_decompiler/sol_complex	74.6±1.39ms	288.6±1.96ms	+286.86%
heimdall_decompiler/sol_simple	1927.8±17.28µs	2.6±0.01ms	+34.87%
heimdall_decompiler/yul_complex	63.2±1.05ms	278.9±7.03ms	+341.30%
heimdall_decompiler/yul_simple	1691.8±43.97µs	2.4±0.04ms	+41.86%
heimdall_disassembler/complex	1269.7±37.45µs	1511.2±22.30µs	+19.02%
heimdall_disassembler/simple	62.0±4.82µs	71.7±4.48µs	+15.65%
heimdall_vm/erc20_transfer	262.7±8.16µs	262.2±8.81µs	-0.19%
heimdall_vm/fib	880.5±38.47µs	868.6±9.90µs	-1.35%
heimdall_vm/ten_thousand_hashes	709.9±19.02ms	730.3±31.65ms	+2.87%

📊 View the full Criterion report