fix(storage): inflated block size estimation caused by shared string buffers

[zhyass]

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

Summary

This PR fixes block size estimation for variable-length data in two places:

1. DataBlock::estimate_block_size() now uses the GC-aware memory size path, so block statistics reflect the effective column footprint after compaction instead of inflated shared-buffer memory.
2. estimated_scalar_repeat_size() is corrected for const variable-length values, so repeated String / Array / Map scalars are no longer underestimated compared with materialized columns.

It addresses two issues:

1. estimated_scalar_repeat_size() could significantly underestimate const variable-length values.
2. estimate_block_size() used a non-GC memory size path, which could overestimate block size for view-backed columns.

Together, these changes make block_size and bytes_uncompressed more stable across insert, recluster, and readback.

After fix

root@localhost:8000/default/default> create or replace table t (
  id int,
  k int,
  s string
) cluster by(k);

root@localhost:8000/default/default> insert into t
select
  number as id,
  number % 10 as k,
  repeat('x', 500000) as s
from numbers(2000);
╭─────────────────────────╮
│ number of rows inserted │
│          UInt64         │
├─────────────────────────┤
│                    2000 │
╰─────────────────────────╯
2000 rows written in 2.576 sec. Processed 2 thousand rows, 953.72 MiB (776.4 rows/s, 370.23 MiB/s)

root@localhost:8000/default/default> insert into t
select
  2000 + number as id,
  number % 10 as k,
  concat('small_', number::string) as s
from numbers(200);
╭─────────────────────────╮
│ number of rows inserted │
│          UInt64         │
├─────────────────────────┤
│                     200 │
╰─────────────────────────╯
200 rows written in 0.154 sec. Processed 200 rows, 6.41 KiB (1.3 thousand rows/s, 41.63 KiB/s)

root@localhost:8000/default/default> alter table t recluster final;
2200 rows written in 4.720 sec. Processed 2.2 thousand rows, 953.73 MiB (466.1 rows/s, 202.06 MiB/s)

root@localhost:8000/default/default> select snapshot_id, block_count, row_count,bytes_uncompressed,bytes_compressed from fuse_snapshot('default','t') limit 2;

╭────────────────────────────────────────────────────────────────────────────────────────────────────╮
│            snapshot_id           │ block_count │ row_count │ bytes_uncompressed │ bytes_compressed │
│              String              │    UInt64   │   UInt64  │       UInt64       │      UInt64      │
├──────────────────────────────────┼─────────────┼───────────┼────────────────────┼──────────────────┤
│ 019d4d39b6a57f9aa6feb8f1d9b9519b │           9 │      2200 │         1000055336 │            19265 │
│ 019d4d399a887e52b7cb4b1a3358a21a │          10 │      2200 │         1000055330 │            13501 │
╰────────────────────────────────────────────────────────────────────────────────────────────────────╯
2 rows read in 0.055 sec. Processed 2 rows, 539 B (36.36 rows/s, 9.57 KiB/s)

Fixes #19658

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 

[zhyass]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5921fdbfc6

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".

[github-actions]

Docker Image for PR

• tag: pr-19657-06b8de6-1775130843

note: this image tag is only available for internal use.

[github-actions]

ClickBench Report

• hits: https://benchmark.databend.com/clickbench/pr/19657/23897146882/hits.html
• internal: https://benchmark.databend.com/clickbench/pr/19657/23897146882/internal.html
• load: https://benchmark.databend.com/clickbench/pr/19657/23897146882/load.html
• tpch100: https://benchmark.databend.com/clickbench/pr/19657/23897146882/tpch100.html
• tpch1000: https://benchmark.databend.com/clickbench/pr/19657/23897146882/tpch1000.html