feat: implement InsertFormatted

[abonander]

Summary

• Adds Client::insert_formatted_with() and InsertFormatted for inserting data in a chosen format with a specified SQL query.
  • InsertFormatted does not buffer data by default, allowing precise control over when the data is sent. For best performance, ensure data is sent in larger chunks or use .buffered() to get BufInsertFormatted which implements buffering.
  • Data may optionally be pre-compressed and buffered separately using CompressedData::new() and InsertFormatted::send_compressed().
  • BufInsertFormatted also implements tokio::io::AsyncWrite for composability.
    • Omitted AsyncWrite on InsertFormatted itself because every write would require copying to a new allocation anyway to send to the connection.

closes #339

Checklist

Delete items not relevant to your PR:

• Unit and integration tests covering the common scenarios were added
• A human-readable description of the changes was provided so that we can include it in CHANGELOG later

Potential additions:

• Client::insert_formatted(), taking a table name and format name
  • Probably best to lump this in with Expose constants or enums for ClickHouse formats #312, actually.
• Method to directly send Bytes, bypassing buffer?
• Rewrite Insert using this API

[slvrtrn]

perhaps then we need to deprecate fetch_bytes, replacing it with fetch_formatted_with for consistency

[abonander]

This could just be insert_bytes_with, I'm not particularly attached to the naming. I just tried to come up with something more descriptive.

[slvrtrn]

I also prefer _formatted_with naming, so it is a good pick!
Let's just keep the naming consistent, i.e., insert_formatted_with + fetch_formatted_with both look good in combination, let's just mark fetch_bytes as deprecated for the next minor release to reduce confusion, and make fetch_formatted or fetch_formatted_with with the same exact content. Also perhaps that new method will be a good candidate to introduce (non-exaustive) enums for common formats, see #312

[abonander]

In this context, the _with suffix is referring to the SQL argument, but on the query side, that would have already been passed to Client::query().

I would call the new method fetch_formatted or fetch_with_format.

[abonander]

Created #369 to follow up because this PR is blocking work on the ADBC driver and I want to avoid scope creep.

[codecov]

Codecov Report

❌ Patch coverage is 87.26790% with 48 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/insert_formatted.rs	87.61%	27 Missing and 14 partials ⚠️
src/insert.rs	78.26%	1 Missing and 4 partials ⚠️
src/request_body.rs	88.23%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!

[abonander]

Ugh, the CodeCov failure is due to 7 lines that I'm not even entirely sure how to hit: https://app.codecov.io/gh/ClickHouse/clickhouse-rs/pull/364?dropdown=coverage&src=pr&el=h1&utm_medium=referral&utm_source=github&utm_content=checks&utm_campaign=pr+comments&utm_term=ClickHouse

We'd have to produce records fast enough to saturate the connection.

There's lines I can try to cover in src/insert_formatted.rs to compensate.