[FEA] Expose TotalBytesRead in Parquet Reader

**Is your feature request related to a problem? Please describe.**

I’d like to propose and discuss adding a metric to the Parquet reader to track the total bytes read from the file during I/O operations. This metric would capture the raw, encoded/compressed bytes transferred from storage/io-source.

The motivation behind this is twofold: first, to provide visibility into the actual data movement measurement at the storage layer, which is helpful for optimizing I/O-bound workflows or profiling. Second, it would enable users to profile and compare the effect of Parquet pruning.

**Describe the solution you'd like**

Currently, `table_metadata` tracks rowgroups but not byte counts.

https://github.com/rapidsai/cudf/blob/a664c1464fef02185350e65a80868ecec59dacb1/cpp/include/cudf/io/types.hpp#L290-L297

Adding a `read_bytes` field to `cpp/include/cudf/io/types.hpp` would align with this existing pattern. The value could come directly from KvikIO, which already tracks raw bytes read. This would require minimal overhead in this book-keeping.