[Veronica0206]

Intrinsic BLS plots substituted covariate-adjusted means into a sharp spline at a shifted knot, so the displayed mean and pointwise intervals could disagree with the fitted RAM model. In the review fixture at time 6, the old curve showed 47.8 while the model expectation was 49.19999935.

This change retains the expansion knot, adjusted internal means, unadjusted slopes and mean knot deviation, and evaluates the fitted smoothed expression with its full joint covariance. It covers standalone, group, mixture, TVC and multivariate growth components; nonintrinsic spline behavior is preserved. Documentation distinguishes their exact-knot interval behavior.

Regression tests compare curves directly with OpenMx expectations and interval propagation with numerical RAM derivatives, including nonzero covariate means, positive/negative knot shifts, simultaneous slope/knot effects and parent-model covariance. Stored BLS interval tests now use the RAM target. NLS, optimizer/retry controls, model-building equations, numerical references and tolerances are unchanged.

Validation at f6ffd1e:

• 552 focused assertions passed, no failures/errors/warnings/skips.
• Independent reproduction: 72 means agree with RAM within 1e-10; derivative and covariance checks pass.
• 688 structural specifications: zero unexplained differences; 76 comparator self-tests passed.
• All 23 stored fits rendered 34 figures without warnings or errors.
• Full local and hosted fitting suites: 5,505 assertions passed each, five documented nonconverged-reference skips, no failures/errors/warnings. Hosted runtime: 2441.692 seconds.
• Declared minimum testthat 3.2.0 on R 4.5.3: 5,324 passed, no skips/failures/errors/warnings.
• All five source-package environments and all four checks of the identical submission archive: zero errors/warnings/notes.
• Portable smoke/linear recovery on three platforms and documentation build passed.
• Extended local check of the exact archive, including the manual and examples: zero errors/warnings, two documented notes (example runtime and local HTML validation tools).

The archive is 2,244,335 bytes, SHA-256 9d1cedea9de1de9ab18676be5edada3bc7987ad3be15c557c9d99e3c75dfe337. Its source revision and 140 included code/test/help/data files match the candidate. Archive validation, package checks, full fitting, portable checks, documentation.

Specified covariance fixtures verify propagation, not empirical nonlinear interval coverage. Existing qualifications for excluded or inadmissible reference fits remain unchanged.

The maintainer will upload the rebuilt v1.0.0 archive to CRAN after validation. This PR does not submit the package or create a release.