Functions should recognize distances given as a symmetric square matrix

[jarioksa]

Some dissimilarity-based functions accept either a "dist" structure or rectangular data from which they calculate the dissimilarities internally. If user supplies the dissimilarities as a symmetric square matrix, this will not be recognized as dissimilarities but is regarded as raw data, and function can calculate dissimilarities of dissimilarities. Several functions already check their input and are able identify symmetric square matrices and handle these as dissimilarities, but many functions assume users do the right thing. This was behind one current issue #147, but it concerns several functions.

The following functions should be fixed:

• adonis
• adonis2 (not in cran-2.3)
• capscale
• dbrda (not in cran-2.3)
• mso (assumes object.xy are coordinates, but can be distances)

The following already do the right thing, but some could still be do it better: anosim (could be cleaned), bioenv.default, metaMDSdist (needs more robust test instead of ==), metaMDS (like previous), monoMDS(cleaner test), mrpp (uses ==), pcnm (does not test symmetry, only squareness).

[jarioksa]

Closed via 3d27f78