Merge of two VNTs can drop results when one contains a PartialArray and the other a full Array

julia> vnt1 = @vnt begin x = [1.0, 2.0] end
VarNamedTuple
└─ x => [1.0, 2.0]

julia> vnt2 = @vnt begin
           @template x = zeros(2)
           x[1] = 3.0
       end
VarNamedTuple
└─ x => PartialArray size=(2,) data::Vector{Float64}
        └─ (1,) => 3.0

julia> merge(vnt1, vnt2)
VarNamedTuple
└─ x => PartialArray size=(2,) data::Vector{Float64}
        └─ (1,) => 3.0

It is a bit unintuitive that x[2] is dropped here. This happens because merge can't line the two up: it just thinks "okay, one is a PA and one is an array and they're different things, so I'll just overwrite the first one". I don't really know what the solution to this is.

Start by reproducing the example with @vnt, PartialArray, full arrays, and merge, then inspect the merge implementation and how these values represent indexed data. Done means merging the shown VNTs preserves both x[1] and x[2], with coverage for the behavior added where the existing merge tests live.