🚧 [DO NOT MERGE] Transcode parquet to cuDF dictionaries for STRING

[mhaseeb123]

Description

This PR adds a prototype implementation of transcoding dictionaries from parquet to cuDF directly when possible and specified for flat string type columns.

What's missing or needs fixing

• Deduplication of dictionary values across row groups. Current prototype just concatenates dictionary values into a single column
• Update the keys to use the deduplicated value indices
• Ensure we aren't just hacking into the decode process with extra kernel launched and actually allocating required buffers and directly decoding data into them in the decode kernel launch.
• Ensure memory and compute patterns are optimal and code modernization
• Support for chunked reading (may need to preserve sub-colchunk dictionaries across sub-passes). Currently, we assume that we will not chunk anything.
• Add more tests

Note: Do not add a merge commit to this PR. Let's keep the commit tree clean.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[mhaseeb123]

Superseded by #22532