Heterogeneous data sometimes detect wrong columns type

There are combinations of heterogeneus data where the wrong column types are discovered.

For example:

using JSONTables
using Tables

nonhomogenous = """
[
    {"a": 1, "b": 2, "c": 3},
    {"b": 4, "c": 8, "d": 5}
]
"""

JSONTables.jsontable(nonhomogenous)

You got:

JSONTables.Table{false, JSON3.Array{JSON3.Object, Base.CodeUnits{UInt8, String}, Vector{UInt64}}}([:a, :b, :c, :d], Dict{Symbol, Type}(:a => Union{Missing, Int64}, :b => Int64, :d => Int64, :c => Int64), JSON3.Object[{
   "a": 1,
   "b": 2,
   "c": 3
}, {
   "b": 4,
   "c": 8,
   "d": 5
}])

d type is detected as Int64 and this throws an error when building a table:

ct = Tables.columntable(jt)
ERROR: MethodError: Cannot `convert` an object of type Missing to an object of type Int64

Start with the supplied nonhomogeneous JSON example and the JSONTables.jsontable call, then run Tables.columntable on the resulting table to reproduce the conversion error. Trace how column types are detected for fields absent from some rows; done means the example builds a table without a Missing-to-Int64 conversion failure.