Condition permutation references on development permute

[gavinsimpson]

Closes #2

Summary

• Regenerate the six affected ANOVA reference objects with development permute.
• Run those reference comparisons only when packageVersion("permute") > "0.9-10".
• Continue running every non-reference assertion with older permute versions.

The ANOVA result tables are unchanged. The references differ only because their nested control attributes include permute's expanded Within and Plots metadata, which now contains the symmetric setting.

Verification

• R CMD check with CRAN permute 0.9-10: OK (324 tinytest results; the six development-only references are skipped)
• R CMD check with development permute based on gavinsimpson/permute@23c9f6a: OK (330 tinytest results), including all six regenerated reference comparisons
• All six statistical result payloads match the existing references after excluding only the dependency-owned control attribute.

[BartJanvanRossum]

see #3