support multiple return values?

I'm trying to slice-broadcast a function that has multiple return values. Ideally I'd collect them into separate arrays. Is this possible witih TensorCast?

Here's a MWE of basically what I'm trying to do:

x = randn(10, 20)
@cast v[k], i[k] := findmax(x[:, k]) 

But this throws the error:

LoadError: don't know what to do with left = (v[k], i[k])

Which is a great error message that leads me to believe I can't do this kind of thing, but I wanted to check.

In this case it's probably not too bad to just collect into a single Vector{Tuple{Float64, Int} and split it afterwards, but my actual example has some more complicated indexing stuff going on that would benefit from some TensorCast magic.

Start by reproducing the Julia MWE using the @cast macro and findmax shown in the issue, then inspect how @cast handles the left-hand side tuple. Define what behavior is required for v[k], i[k] and verify that the expression can collect the multiple return values into separate arrays without breaking existing slice-broadcasting cases.