sql: add ClickHouse Map type support for sql_insert

[rickysaltzer]

The ClickHouse driver requires map arguments to have concrete Go types matching the column schema (e.g. map[string]int64), but Bloblang args_mapping produces generic map[string]any values. This causes inserts to fail for any column using ClickHouse Map types.

Introduce an argsConverter for ClickHouse that introspects column scan types at connect time via a SELECT ... LIMIT 0 query, then normalizes map and slice arguments through a JSON round-trip into the expected Go types before passing them to the driver.

Applied to both sql_insert output and processor components. Includes integration tests covering common map types.

[CLAassistant]

All committers have signed the CLA.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[rickysaltzer]

In reference to #3819

[rickysaltzer]

Full disclosure, I used Claude to assist with this patch as I don't normally write in Go. Feel free to make any changes to conform to the project's standards.

[josephwoodward]

Hi @rickysaltzer, thank you for your pull request. There's a number of linter violations that would be good to clean up and I've added a few comments.

If you are using Claude then you can also use the Claude skill here to review the pull request which can be pretty helpful.

[rickysaltzer]

Hey @josephwoodward -

Thanks for such a quick review, I've sent in some changes to hopefully address your comments.

[rickysaltzer]

Commits

1. 6475aea "Address Comments" — message format violation: should use sql: prefix (e.g., sql: remove unused regex and add normalization failure logging). Message is vague and doesn't describe the change. This is a fixup to the first commit and should be squashed before merge.

Review
Code changes look good. The argsConverter extension point is clean, reflection/JSON round-trip logic is correct, no race conditions or resource leaks, and integration tests follow existing patterns. The three issues flagged by @josephwoodward in the prior review round (unused parseInsertTableColumns/insertRegexp, unnecessary UseNumber(), silent normalization failures) appear to have been addressed in commit 6475aea.

LGTM once the commits are squashed.

[josephwoodward]

Thanks for addressing the comments @rickysaltzer. I see the linter is complaining about something, if you wouldn't mind addressing that when you have a moment that'd be great. Also, if you rebase your changes on the main branch the tests should go green.

[rickysaltzer]

Hey @josephwoodward - linting should be taken care of if you want to approve the workflows to check.

[josephwoodward]

[josephwoodward]

Thank you for your contribution @rickysaltzer, your change is now available in v4.86.0