feat(benchmark): add format-level Parquet read/write micro-benchmarks

Add paimon-parquet-format-benchmark, driving ParquetWriterBuilder and
ParquetFileBatchReader directly so a change confined to the format layer can
be attributed without the catalog lookup, split planning, merge/sort and
commit that the table-level read_write_benchmark includes.

Writer cases cover flat primitives, VARCHAR by cardinality with and without
file-level dictionary encoding, dictionary-encoded VARCHAR/INTEGER input
arrays against flat controls holding value and width fixed, DECIMAL at
precision 9/18/38, nested STRUCT/LIST/VECTOR/MAP, null density, rows per
AddBatch call, column count, row group size, the byte-triggered writer memory
threshold, and every codec Parquet accepts.

Reader cases cover full scan, projection, predicate-filtered reads with
page-index filtering on and off, skip-heavy reads driven by a strided
selection bitmap whose strides straddle the coalesce hole limit, null
density, DECIMAL, DOUBLE, dictionary against plain files, rows per NextBatch
call, and nested column reads.

Each case reports ns_per_row and bytes_per_row next to ns_per_input_row and
bytes_per_input_row, so settings that prune by different amounts stay
comparable, plus the reader's own row-group and batch counters. Read cases
assert on the rows they materialized and the process exits non-zero on any
SkipWithError, so a case that stops measuring what it claims fails instead of
looking fast.

Every axis is swept on its own rather than as a combined matrix, so each case
answers one question and a change can be attributed to it.

Also add paimon-parquet-format-benchmark-test, which guards the format-layer
assumptions the benchmark rests on. The benchmark only builds under
PAIMON_BUILD_BENCHMARKS, which CI does not set, so the test is what keeps
codec acceptance, dictionary round-trips, DECIMAL precision, nested types,
predicate and bitmap row counts, encoding agreement and the memory-threshold
flush covered.

Closes #228

Author: SteNicholas

benchmark/CMakeLists.txt

@@ -63,6 +63,20 @@ if(PAIMON_BUILD_BENCHMARKS)
                          ${PAIMON_BENCHMARK_LINK_TOOLCHAIN}
                          EXTRA_INCLUDES
                          ${CMAKE_SOURCE_DIR})
+
+    add_paimon_benchmark(parquet_format_benchmark
+                         SOURCES
+                         parquet_format_benchmark.cpp
+                         STATIC_LINK_LIBS
+                         arrow
+                         parquet
+                         ${PAIMON_BENCHMARK_STATIC_LINK_LIBS}
+                         test_utils_static
+                         Threads::Threads
+                         ${PAIMON_BENCHMARK_PLATFORM_LINK_LIBS}
+                         ${PAIMON_BENCHMARK_LINK_TOOLCHAIN}
+                         EXTRA_INCLUDES
+                         ${CMAKE_SOURCE_DIR})
 endif()
 
 if(PAIMON_BUILD_TESTS)
@@ -74,4 +88,21 @@ if(PAIMON_BUILD_TESTS)
                     STATIC_LINK_LIBS
                     paimon_shared
                     ${GTEST_LINK_TOOLCHAIN})
+
+    # Guards the format-layer assumptions parquet_format_benchmark.cpp is built on. The benchmark
+    # itself is only compiled under PAIMON_BUILD_BENCHMARKS, which CI does not set, so this test is
+    # what keeps those assumptions covered.
+    add_paimon_test(parquet_format_benchmark_test
+                    SOURCES
+                    parquet_format_benchmark_test.cpp
+                    EXTRA_INCLUDES
+                    ${CMAKE_SOURCE_DIR}
+                    STATIC_LINK_LIBS
+                    arrow
+                    parquet
+                    paimon_shared
+                    ${PAIMON_LOCAL_FILE_SYSTEM_SHARED_LINK_LIBS}
+                    ${PAIMON_PARQUET_FILE_FORMAT_STATIC_LINK_LIBS}
+                    test_utils_static
+                    ${GTEST_LINK_TOOLCHAIN})
 endif()