refactor: new impl for loading CSV.

[youngsofun]

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

Summary

the original InputContex is intended to include all kinds of file formats, and both streaming load and copy, which turn to be too complicated to maintain, and there are many differences in nature between parquet and text files.
this pr introduce new pipeline for text, row-based formats, and force on copy.
the code is mainly in databend-common-storages-stage::read::row_based
CSV format is refactored with this new pipeline firstly.

logically this is a refactor, but at code level many codes are rewritten (some details are refined by the way). So Add new setting enable_new_copy_for_text_formats.
if there are some problem with new impl, user can fallback to the old one.
because there are already many tests for CSV, I set enable_new_copy_for_text_formats=1 by default.

• Fixes #[Link the issue here]

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 

[youngsofun]

@b41sh thank you very much! please review again (take your time, this PR is not in hurry).

[sundy-li]

Do you have any performance results with the old impl ?

You can use copy into table with null engine to test the performance.

[youngsofun]

Do you have any performance results with the old impl ?

You can use copy into table with null engine to test the performance.

ok, I will try it later.

as a refactor, the only place that may lead to diff in performance is the reading of file data (less batch buffered).
If there really is a difference in this aspect, it would only become apparent when running on slower storage.

[youngsofun]

@sundy-li

A preliminary test indicates negligible performance difference (less than 1%).

Settings:

A single CSV file of 700MB with two integer columns
Tested in both compressed and uncompressed formats.

--

this test is too cpu bounded, add another:

• one column with a 100-char string
• add max_threads = 1 and 10

not diff either.