CRAN release 2.7-2

vegan 2.7-2 is on CRAN. Many files have changed, but most of these changes are small technical tweaks that do not influence most users (but may make vegan marginally faster and more robust in some rare cases). Two most important changes are:

• permutation tests for partial RDA and db-RDA adopt similar pattern as partial CCA in CRAN release 2.6-6, and use "residualized predictors". Non-partial RDA and db-RDA are not influenced. Earlier add1 method and anova(..., by = "axis") internally used partial model, but this does not make sense with residualized permutations, and were redesigned. Now add1 and drop1 are consistent for ordination models, whereas they used different internal models in earlier releases.
• The working with tibbles is more robust. Earlier vegan failes its tests if all example data frames were replaced with tibbles. The main problem was that tibbles often do not have real rownames, but vegan assumed named results. Now tibbles and data frames give similar results also for names.