Wrong answer when input to findColumn is spread across multiple buffers

[wolffcm]

This test case will currently fail:

testcase findColumnWithGroup {
    // The result of grouping and then ungrouping
    // will be for a single table to be spread over
    // four buffers.
    input = array.from(rows: [
        {m: "m", k: "north", v: 10, _time: 2020-02-20T00:00:00Z},
        {m: "m", k: "south", v: 20, _time: 2020-02-20T00:00:00Z},
        {m: "m", k: "east", v: 30, _time: 2020-02-20T00:00:00Z},
        {m: "m", k: "west", v: 40, _time: 2020-02-20T00:00:00Z},
    ])
       |> group(columns: ["k"])
       |> range(start: -100y)
       |> group()

    arr = input |> findColumn(fn: (key) => true, column: "v")
    got = array.from(rows: [{val: display(v: arr)}])
    want = array.from(rows: [{val: display(v: [10, 20, 30, 40])}])
    testing.diff(got, want)
}

It fails like this on current master:

#  _diff=string,val=string
_diff=-,val=[
    10,
    20,
    30,
    40
]
_diff=+,val=[40]

That is, the result should be [10, 20, 30, 40] but it is [40]. This happens when the table input to findColumn and related functions is spread across multiple buffers which will happen in the query above. In this case, rather than concatenating the values in each row, it will simple take the single value in the last buffer.