feat: Implement columnStatistics() for Nimble SelectiveNimbleReader to enable file-level filter pushdown

[kewang1024]

Summary:
Nimble writes rich file-level column statistics via VectorizedFileStats
(min/max/count/nullCount per column for integer, floating-point, and
string types) in the "columnar.vectorized_stats" optional section.
However, SelectiveNimbleReader::columnStatistics() returns nullptr,
which means Nimble files cannot participate in file-level filter
pushdown — the mechanism used in HiveConnectorUtil::testFilters() to
skip entire files whose stats prove no rows can match the query filter.

This diff bridges the gap by implementing columnStatistics() in
SelectiveNimbleReader:

• Adds toCommonColumnStatistics() helper that converts
  nimble::ColumnStatistics to dwio::common::ColumnStatistics subclasses:
  • IntegralStatistics -> IntegerColumnStatistics (min/max)
  • FloatingPointStatistics -> DoubleColumnStatistics (min/max)
  • StringStatistics -> StringColumnStatistics (min/max)
  • Base ColumnStatistics -> base ColumnStatistics (valueCount/hasNull/size)
• Loads VectorizedFileStats in ReaderBase at construction time,
  exposed via fileColumnStats(). This is shared by both columnStatistics()
  (for file-level filter pushdown) and computeStatsBasedRowSize()
  (for row size estimation), eliminating duplicate stats loading.

End-to-end call chain for file-level filter pushdown:

Query with filter WHERE col > 200 on a Nimble file with col values [0, 100]:

SplitReader::prepareSplit()
  -> checkIfSplitIsEmpty()
    -> filterOnStats()
      -> testFilters()
        -> reader->columnStatistics(colId)
          -> [NEW] ReaderBase::fileColumnStats() (loaded at construction)
          -> [NEW] toCommonColumnStatistics() converts to IntegerColumnStatistics{min=0, max=100}
        -> testFilter(filter=">200", stats={min=0, max=100}, ...)
          -> testInt64Range(0, 100, mayHaveNull) returns false
        -> return false -> FILE SKIPPED

Differential Revision: D98945345

[meta-codesync]

@kewang1024 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D98945345.