Expand Scientific Regression Coverage for SIST 1.0.0

[harryswift01]

Summary

This PR will expand the SIST scientific regression suite to cover all supported calculation modes for the 1.0.0 release.

The existing competition regression coverage has been extended with maintained baselines for melting, Z-DNA, and cruciform calculations so that future numerical drift can be detected across the supported scientific workflows.

Changes

Scientific regression baselines:

• Add maintained 1.0.0 reference outputs for melting, Z-DNA, and cruciform calculations.
• Retain the existing competition reference outputs and historical inherited output for provenance.
• Document the supported scientific regression baselines and how reference updates should be handled.

Shared SIST test execution:

• Generalise the test fixture infrastructure so all supported SIST calculation modes use the same execution path.
• Add parameterised execution for melting, Z-DNA, and cruciform calculations.
• Reuse the same controlled input sequence and temporary maintained-source build for all regression cases.

Regression and CLI coverage:

• Add CLI checks for the individual transition calculations.
• Compare deterministic scientific metrics against the versioned 1.0.0 baselines.
• Compare every reported sequence position, base, transition probability, and energy value where applicable.
• Continue comparing competition P_melt, P_Z, and P_cruciform values across the full sequence.
• Ignore only the variable Run time value during scientific regression comparison.

Impact

• All supported SIST scientific calculation modes are now covered by regression tests.
• Even small changes to deterministic scientific output should cause the regression suite to fail.
• Establishes a versioned 1.0.0 scientific baseline for future maintenance and development.
• Reduces the risk of unintended numerical changes being introduced without detection.

[jimboid]

Greater coverage of codebase by tests is a welcome addition. Do you have a feel for how the maxwindow change yesterday would have passed these had that change been made after this?

[harryswift01]

Greater coverage of codebase by tests is a welcome addition. Do you have a feel for how the maxwindow change yesterday would have passed these had that change been made after this?

Yeah I don’t think this should affect the calculation itself, as the warning was coming from MinWindowSize and MaxWindowSize being used in the constructor before they had been initialised, and the fix just changes that initialisation loop to use the fixed array size instead. I have also just done a quick check by restoring the old behaviour and running the new regression suite against it, and all 16 tests still passed, then restored the fix and they all passed again, so that gives me a bit more confidence that this change isn’t affecting the actual calculations.