DataContainers only stores a deep copy of the data when !data_are_columns

[ilopezgp]

We currently have

struct DataContainer{FT <: Real}
     #stored data, each piece of data is a column [data dimension × number samples]
     stored_data::AbstractMatrix{FT}
     #constructor with 2D arrays
     function DataContainer(stored_data::AbstractMatrix{FT}; data_are_columns = true) where {FT <: Real}

         if data_are_columns
             new{FT}(stored_data)
         else
             new{FT}(permutedims(stored_data, (2, 1)))
         end
     end
 end

The behavior for this struct is as follows,

julia > u = rand(5,3)

output > 5×3 Matrix{Float64}:
 0.553978   0.100353  0.455385
 0.393232   0.138405  0.74255
 0.0660367  0.174017  0.401449
 0.669909   0.231632  0.249801
 0.113858   0.4794    0.315958

julia > data1 = DataContainer(u)
julia> data2 = DataContainer(u, data_are_columns = false)

julia> get_data(data1)

output > 5×3 Matrix{Float64}:
 0.273102   0.24338   0.728426
 0.0066842  0.320747  0.562981
 0.420169   0.521042  0.620609
 0.431041   0.177357  0.631251
 0.284572   0.120433  0.876362

julia> u[1,1] = -5

julia> get_data(data1)

output > 5×3 Matrix{Float64}:
 -5.0        0.24338   0.728426
  0.0066842  0.320747  0.562981
  0.420169   0.521042  0.620609
  0.431041   0.177357  0.631251
  0.284572   0.120433  0.876362

# With data_are_columns, the memory is shared.

julia> get_data(data2)

3×5 Matrix{Float64}:
  0.273102  0.0066842  0.420169  0.431041  0.284572
  0.24338   0.320747   0.521042  0.177357  0.120433
  0.728426  0.562981   0.620609  0.631251  0.876362

# With !data_are_columns, the memory is not shared.

[odunbar]

Yeah this is unexpected... let's always make a copy on setting.