relplot does not handle numpy-types for dimensional variables

[mwaskom]

Test case:

sns.relplot(
    data=tips,
    x="total_bill",
    y=tips["tip"].to_numpy(),
    col=tips["time"].to_numpy(),
)

Note how it handles y= fine*, but does not create two columns (or error/warn in any useful way).

displot handles this better:

sns.displot(
    data=tips,
    x="total_bill",
    y=tips["tip"].to_numpy(),
    col=tips["time"].to_numpy(),
)

displot and replot solve the problem of initializing a FacetGrid from vector data differently. I do not remember if this ever worked in relplot and is a regression (quite possibly not) and, if not, whether that was a failure of implementation or the result of a decision to punt on a particularly tricky issue. If the latter, it should at least give feedback about why it is not working. It looks like this never worked.

• the internal name used for y here, _y, shows up in the y label, which I'm also not sure that we want. Note that there is no y axis label for the displot, although the internal name for the column variable is used because FacetGrid titles include that in their title template.