RunEndEncodeTableColumns doesn't change the table's schema type to reflect that the column is run-end-encoded

[lesterfan]

Describe the bug, including details regarding any error messages, version, and platform.

This is half a bug report regarding the RunEndEncodeTableColumns gtest util and half a usage question.

If a string column in an arrow::Table is run-end encoded, should the corresponding schema type be arrow::utf8() or arrow::run_end_encoded(arrow::int32(), arrow::utf8())? In the current main branch, the RunEndEncodeTableColumns gtest util returns a table like

ree_table = col: string
----
col:
  [

    -- run_ends:
      [
        1,
        2,
        3,
        4
      ]
    -- values:
      [
        "a",
        "b",
        "c",
        "d"
      ]
  ]

whereas I would have expected a table like

ree_table = col: run_end_encoded<run_ends: int32, values: string>
  child 0, run_ends: int32 not null
  child 1, values: string
----
col:
  [

    -- run_ends:
      [
        1,
        2,
        3,
        4
      ]
    -- values:
      [
        "a",
        "b",
        "c",
        "d"
      ]
  ]

I'm not sure what the right behavior is here. My instinct is that the second is more correct since I see in the codebase that certain features are disabled for run-end-encoded types (example), so we would want the schema to be accurate to reflect what the library currently supports on the column. I'm not sure whether the existence of these unsupported features implies that having a plain string type in the schema is incorrect though. I definitely don't have a lot of context here though, so let me know if I'm missing something in my considerations here 🙂

Component(s)

C++

[lesterfan]

I put up #45535 to change this function to also modify the schema type pending further discussion here.