ncplane_resize_internal() ought realloc when x dimension doesn't change

[dankamongmen]

See #2323. ncplane_resize_internal() ought be optimized for the case where the x dimension doesn't change -- in this case, we want to realloc() if possible, since we needn't restripe the existing data (when the x dimension does change, we ought go ahead and malloc()+copy, as we'd otherwise realloc()+restripe, where the realloc() might require a copy itself).

this is an important optimization if someone's laying out a plane that grows downward with the content. this way, we only require a copy when we actually must move (i.e. within realloc()).

[dankamongmen]

this is less trivial than it might sound...we allocate the new framebuffer early in ncplane_resize_internal(), and it's a loooong function...

also, we can only do this when the old dimx equals the new dimx, and it is equal to keeplenx.

[dankamongmen]

we also ought do this whenever we're not keeping anything, to avoid a potential copy there.

[dankamongmen]

alright, i think i've got it working...

[dankamongmen]

yeah, we've got this in there now.