[BUG] Memcheck error cudf::io::parquet::detail::<unnamed>::gpuComputeStringPageBounds<unsigned char>()

[davidwendt]

Describe the bug
Nightly memcheck build is reporting out-of-bounds read error in cudf::io::parquet::detail::<unnamed>::gpuComputeStringPageBounds<unsigned char>() when running gtest ParquetChunkedReaderTest.TestChunkedReadWithString

[ RUN      ] ParquetChunkedReaderTest.TestChunkedReadWithString
========= Invalid __global__ read of size 8 bytes
=========     at 0x3a50 in void cudf::io::parquet::detail::<unnamed>::gpuComputeStringPageBounds<unsigned char>(cudf::io::parquet::detail::PageInfo *, cudf::device_span<const cudf::io::parquet::detail::ColumnChunkDesc, (unsigned long)18446744073709551615>, unsigned long, unsigned long)
=========     by thread (0,0,0) in block (5,0,0)
=========     Address 0x7f04727bcc08 is out of bounds
=========     and is 489 bytes after the nearest allocation at 0x7f04727bca00 of size 32 bytes
=========     Saved host backtrace up to driver entry point at kernel launch time
=========     Host Frame: [0x3344e0]
=========                in /usr/lib/x86_64-linux-gnu/libcuda.so.1
=========     Host Frame: [0x1488c]
=========                in /conda/envs/rapids/lib/libcudart.so.11.0
=========     Host Frame:cudaLaunchKernel [0x6c318]
=========                in /conda/envs/rapids/lib/libcudart.so.11.0
=========     Host Frame:cudf::io::parquet::detail::ComputePageStringSizes(cudf::detail::hostdevice_vector<cudf::io::parquet::detail::PageInfo>&, cudf::detail::hostdevice_vector<cudf::io::parquet::detail::ColumnChunkDesc> const&, rmm::device_uvector<unsigned char>&, unsigned long, unsigned long, int, unsigned int, rmm::cuda_stream_view) [0x14fcd2d]
=========                in /conda/envs/rapids/lib/libcudf.so
=========     Host Frame:cudf::io::parquet::detail::reader::impl::decode_page_data(unsigned long, unsigned long) [0x151fe2a]
=========                in /conda/envs/rapids/lib/libcudf.so
=========     Host Frame:cudf::io::parquet::detail::reader::impl::read_chunk_internal(bool, std::optional<std::reference_wrapper<cudf::ast::expression const> >) [0x15234d8]
=========                in /conda/envs/rapids/lib/libcudf.so
=========     Host Frame:cudf::io::parquet::detail::reader::impl::read_chunk() [0x1524292]
=========                in /conda/envs/rapids/lib/libcudf.so
=========     Host Frame:cudf::io::parquet::detail::chunked_reader::read_chunk() const [0x1513303]
=========                in /conda/envs/rapids/lib/libcudf.so
=========     Host Frame:cudf::io::chunked_parquet_reader::read_chunk() const [0x1400f06]
=========                in /conda/envs/rapids/lib/libcudf.so
=========     Host Frame:(anonymous namespace)::chunked_read(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, unsigned long, unsigned long) [clone .constprop.0] [0x2d1545]
=========                in /cudf/cpp/build/gtests/PARQUET_TEST
=========     Host Frame:ParquetChunkedReaderTest_TestChunkedReadWithString_Test::TestBody() [0x2d6311]
=========                in /cudf/cpp/build/gtests/PARQUET_TEST

Steps/Code to reproduce bug
The following can be used to reproduce the error.

compute-sanitizer --tool memcheck gtests/PARQUET_TEST --gtest_filter=ParquetChunkedReaderTest.TestChunkedReadWithString --rmm_mode=cuda

This error may have been introduced sometime in 23.12 but was masked by the memcheck failures due to #14440 which was only recently resolved in 24.02.
@etseidl @nvdbaranec @vuule

[etseidl]

Thanks @davidwendt. On it...