Query Fails on columns of type Any

df = DataFrame(id=1:3,p=[[1, missing],[2, 3],[3, 4,missing]])

@from i in df begin
    @from j in i.p
    @select {i.id,p=j}
    @collect DataFrame
end

the above works because the column id has type Int, and column p has type Vector{Array{Union{Missing, Int64},1}}, but it fails on

df = DataFrame(id=1:3,p=Any[[1, missing],[2, 3],[3, 4,missing]])

This seems to me a bug, because the result DataFrame of above query have all the columns having type Any, preventing further query operations on the query result.

code tested on Query master v0.12.3-DEV, Julia v1.4.2 on windows 10.

Start by running the two Julia reproductions from the issue on Query master v0.12.3-DEV, comparing the query result types for the DataFrame column declared normally and the one declared as Any. Trace the query materialization and type inference path to identify why Any columns prevent further queries. Done means the Any-column case produces a usable query result without regressing the typed-column case.