`DataFrame#to_wide` returns wrong wide table when rows are not ordered

[heronshoes]

DataFrame#to_wide returns wrong wide table when rows are not ordered.

wide = DataFrame.new(
  rows: %w[row1 row2],
  Column1: [1.1, 1.2],
  Column2: [2.1, 2.2]
)
wide
=>
#<RedAmber::DataFrame : 2 x 3 Vectors, 0x00000000004422e4>
  rows      Column1  Column2
  <string> <double> <double>
0 row1          1.1      2.1
1 row2          1.2      2.2

long = wide.to_long(:rows)

=> 
#<RedAmber::DataFrame : 4 x 3 Vectors, 0x00000000004760f8>
  rows     NAME        VALUE
  <string> <string> <double>
0 row1     Column1       1.1
1 row1     Column2       2.1
2 row2     Column2       2.2
3 row2     Column1       1.2

long.to_wide == wide

=>
true

shuffled = long.shuffle
shuffle
=> 
#<RedAmber::DataFrame : 4 x 3 Vectors, 0x0000000000516850>
  rows     NAME        VALUE
  <string> <string> <double>
0 row1     Column1       1.1
1 row2     Column2       2.2
2 row2     Column1       1.2
3 row1     Column2       2.1

shuffled.to_wide
=> 
#<RedAmber::DataFrame : 2 x 3 Vectors, 0x000000000052ec84>
  rows      Column1  Column2
  <string> <double> <double>
0 row1          1.1      2.1
1 row2          2.2      1.2

shuffled.to_wide == wide
=>
false