PR: feat: make structural enumeration the CLI verify default #42

[metasmile]

Summary

Integrates the structural enumeration pipeline (issue #40, tagma_core CoordSpace) into the ev verify CLI path, closing issue #42 and unlocking the two speed-ignored CLI tests.

Changes

• src/verify/compose.rs: raw_total_combinations computes the cartesian product size from field domains without enumeration, sharing the overflow and MAX_COMBINATIONS guard with expand_all.
• src/verify/evaluate.rs: evaluate_structural returns the raw total plus the evaluations of the structurally valid subset (StructuralEnum + evaluate_all), with an invariant guard that fails loudly if the emitted count exceeds the raw total.
• src/report/reporter.rs: ReporterCapable::report gains a total parameter; the four reporters derive failed = total - passed via split_counts, which asserts passed <= total. The TextReporter prints the failure header only when failure rows are present.
• src/main.rs: verify routes through evaluate_structural; the simulate channel is unchanged.
• Tests: the two ignored CLI tests now run and assert summary counts (16,384 / 2,560 / 13,824 R4; 33,554,432 / 196,608 / 33,357,824 full). Three lib-level tests in tests/structural_enum.rs pin evaluate_structural against expand_all + evaluate_all, including an enable_mask differential test (alu_ext).
• run.sh: the 33M fixture is asserted in CI instead of skipped; CVA6 full and R4 counts checked via _verify_check.
• Docs: devlog docs/devlog/2026-08-23-cli-structural-integration.md, README pipeline and validation table updates.

Measured result

Metric	Before	After
ev verify CVA6 full (33.5M raw)	10.4 s	0.21 s
Evaluation rows for CVA6 full	33,554,432	196,608 (valid subset)
CLI tests ignored for speed	2	0

Counts are unchanged on every committed fixture (verified across all 10 fixtures). The 33M CLI test runs in 0.45 s in debug.

Validation

• cargo test --release: 97 passing (73 lib + 16 CLI + 8 structural), none ignored.
• bash run.sh --code: fmt, clippy, build, test, bench all pass.

Related

• Issue CLI structural integration + CVA6 full struct_enum benchmark verification #42 (this PR), issue Integrate synTagma CoordSpace: replace Vec-based domain expansion with structural coordinate space #40 (merged via PR feat: integrate synTagma CoordSpace — structural enumeration replaces Vec-based expansion (#40) #41).
• Note: overlaps tests/cli_test.rs, run.sh, and the README validation table with the PR for Tagma hw RTL fixture specs: feasible verification YAML specs and plan #47 (Tagma fixtures); the R4 test conflict resolves to this PR's version.

[metasmile]

Reviewed the reporter refactor. One actionable item: the "and N more failures" count should reflect the present rows in the structural pipeline. The split_counts invariant approach is sound.

[metasmile]

The invariant is enforced with assert!, which panics. The trait returns bool rather than Result, so a panic is the loudest available signal for a caller-side contract violation, and it is preferable to a silent usize wrap in release. This is a reasonable design choice; documenting it in the trait doc comment (already done) is sufficient.

[metasmile]

The "and N more failures" count is failed_count - MAX_PRINTED_FAILURES, where failed_count includes combinations that were never enumerated in the structural pipeline (the structurally absent ones). This overstates the unprinted detail. Since failed_rows holds the present rows, failed_rows.len() - MAX_PRINTED_FAILURES is the accurate count of remaining listed rows; on the full expansion path the two are identical.