feat: enable Decimal64 handling in fuse table deserialization

[dantengsky]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

• Fuse-table parquet deserialization now loads decimal columns with precision ≤18 as Decimal64 instead of Decimal128.

  Parquet already picks the physical type based on precision during serialization, so only the read path needed to change. This remains fully backward compatible and safe for rolling upgrades.

• Fuse-table native serialization and deserialization now store Decimal64 columns as true 64-bit values per row instead of widening them to 128 bits.

  This change is not backward compatible: tables produced before this PR cannot be read by a build that includes this PR and will fail during Decimal64 decoding.

  But given the clear performance gain and the native format still being in the research phase, accepting this compatibility trade-off seems reasonable.

• To keep rolling upgrades viable, we leave the table-column metadata protobuf in meta untouched even though it still lacks Decimal64; changing it would break older nodes in mixed-version deployments.

  Technically we could push part of the Decimal conversion logic into schema_from_to_protobuf_impl.rs, but that would force edits to existing proto_conv unit tests which would break that module’s append-only testing style and feels like a poor trade-off.

Performance

• Environment: r7i.8xlarge (single node), disk cache on, TPC-H SF1000 lineitem fully cached on disk.
• Builds: This PR vs v1.2.848-nightly.
• Queries
  • q1: select l_quantity from lineitem ignore_result
  • q2: select l_extendedprice from lineitem ignore_result

Three hot runs per query/version (all values in seconds).

Query	Version	Run 1	Run 2	Run 3	Avg
q1	This PR	1.04	1.02	1.00	1.02
	v1.2.848-nightly	3.28	3.47	3.32	3.36
q2	This PR	4.43	4.37	4.38	4.39
	v1.2.848-nightly	6.29	6.23	6.29	6.27

---

• fixes: Enable Decimal64 handling throughout the Fuse-table Parquet deserialization pipeline. #18993

Tests

• Unit Test
• Logic Test
• Benchmark Test
• No Test - Explain why

Type of change

• Bug Fix (non-breaking change which fixes an issue)
• New Feature (non-breaking change which adds functionality)
• Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• Documentation Update
• Refactoring
• Performance Improvement
• Other (please describe):

---

This change is 

[dantengsky]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".