keys return LinearIndices

[theogf]

When calling keys on a TupleVector one gets

julia> x = TupleVector((;x=randn(10), y=randn(10)))
10-element TupleVector with schema (x = Float64, y = Float64)
(x = 0.1±1.1, y = 0.18±0.82)

julia> keys(x)
10-element LinearIndices{1, Tuple{Base.OneTo{Int64}}}:
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
julia> keys((;x=1, y=2))
(:x, :y)

when I would expect (:x, :y) I suppose?
Happy to make a PR if you agree that it should be the correct behaviour

[cscherrer]

I see what you mean here, my only concern is that TupleVector <: AbstractVector, and some people might expect generic methods to behave as they do for vectors. We should consider how this is handled in StructArrays.jl and DataFrames.jl. More generally, we could look at ArrayInterface.jl and Tables.jl and try to find some commonality -- nice if we could conform to both.