Fix hybrid scan parquet reader incorrectly parse bloom filter - #22901

…; remove outdated test for Bloom filters with real data

Need to rebuild all the code for A/B test

… read_parquet

Updated the test for hybrid scan bloom filtering to explicitly compare its behavior against the read_parquet method. Enhanced debug logging to provide detailed output of the bloom filter data being processed, ensuring consistency in row group retention between the two methods. Adjusted comments and structure for better readability and understanding of the test's purpose.

Refactored the hybrid scan bloom filter test to utilize parameterized fixtures for improved flexibility and clarity. This change allows for multiple test cases to be run with different input values, enhancing the robustness of the test against various scenarios. Updated the test structure and comments for better readability and understanding of the A/B comparison with the read_parquet method.

…and clarity

Updated the hybrid scan implementation to replace the previous method of fetching bloom filter data with a new function, `fetch_bloom_filters_to_device_async`, which strips headers and ensures 32-byte alignment. This change enhances the efficiency of bloom filter processing and simplifies the codebase. Additionally, updated related tests to reflect these changes and ensure consistency in behavior across different methods of bloom filter handling.

Modified copyright statements in multiple files to include "NVIDIA CORPORATION & AFFILIATES." Enhanced code readability by breaking long parameter descriptions into multiple lines in the `parquet_io_utils.hpp` file. Removed outdated debug logging and unnecessary comments in the `bloom_filter_reader.cu` and `parquet_io_utils.cpp` files to streamline the codebase.

…vice_async_impl`

Simplified comments and improved code clarity by removing outdated explanations regarding bloom filter header handling. The changes focus on enhancing readability while maintaining the functionality of fetching bloom filter data to the device. This aligns with previous efforts to streamline the hybrid scan implementation and improve performance.

…gement

- Removed unnecessary `copy_dsts` vector and related null pointer handling.
- Simplified the logic for populating `copy_srcs` and `copy_sizes`.
- Ensured that device memory allocation is only performed when `device_base` is not null.
- Improved clarity and maintainability of the code by restructuring deferred read handling.

This change enhances the efficiency of the bloom filter fetching process in the Parquet I/O utilities.