Yes, it's a little late here ...

I've made this a pending deprecation, which feels a bit better than a very long deprecation period.

--> Means we'll still have to deprecate explicitly in a later version.

PendingDeprecationWarning was not playing nicely with sphinx (warnings causing sphinx to fail). I suppose we'd have to include special handing for that.

Instead of aPendingDeprecationWarning I went back to only discouraging this code. In both cases we'd have to add aDeprecationWarning to later releases if we want to finally remove it. So there's no real benefit of the pending deprecation warning.

Macos test failure seems unrelated (subprocess timed out).

@dstansby I think that CI is happier now...