What changes were proposed in this pull request?

(Fixes: #8623)

This PR supports file meta for all supported format and row index for parquet.

Supporting File Meta

To support File Meta, FileReaderWrapper is renamed to BaseReader, and add a member named DB::Columns addVirtualColumn(DB::Chunk dataChunk, size_t rowNum = 0) const, which is called at XXFileReader::pull.

               ┌──────────────┐
               │  BaseReader  │
               └──────┬───────┘
                      ˄
                      │
          ┌───────────┴───────────────┐
          │                           │
┌──────────────────────┐     ┌───────────────────┐
│ConstColumnsFileReader│     │ NormalFileReader  │
└──────────────────────┘     └───────────────────┘

NormalFileReader::pull is responsible for reading real data from file, and ConstColumnsFileReader::pull is responsible for generating n rows of meta data when there is no need to read real data.

After read data from file, file meta are added.

Supporting Row index for parquetParquetInputFormat::generate

To understand what a row index is, please refer to this #37228. and VirtualColumnRowIndexReader

To support row index for parquet, I refactor FormatFile::InputFormat and create a new child class ParquetInputFormat

┌───────────────┐
│  InputFormat  │
└──────┬────────┘
       ˄
       │
┌──────┴─────────────┐
│ ParquetInputFormat │
└────────────────────┘

In ParquetInputFormat::generate, we did same as XXFileReader::pull, reading real data from parquet file first, and then add row index.

How was this patch tested?

spark 35 test are added

  enableSuite[GlutenParquetFileMetadataStructRowIndexSuite]
  enableSuite[GlutenParquetRowIndexSuite]
    .excludeByPrefix("row index generation")
    .excludeByPrefix("invalid row index column type")

see https://opencicd.kyligence.com/blue/rest/organizations/jenkins/pipelines/gluten/pipelines/gluten-ci/runs/14511/nodes/151/steps/205/log/?start=0