feat!: Move to Arrow implementation

[hermanschaaf]

• JSON
• CSV
• Parquet

In the parquet writer, we are converting extension columns to String. This is how it was also implemented before. One difference, however, is that UUID columns are now delimited with dashes, where before they were not. Timestamps now also use the default Arrow format, not the one we were using previously.

For these reasons, and because this is a fairly major internal change, we have decided to make this a breaking change.

Depends on cloudquery/plugin-sdk#783

[hermanschaaf]

I still need to do a little bit more clean-up to fix conflicts in the go.mod and also create PRs for other repos (notably plugin-sdk) with changes that were necessary to make this work.

Mostly updated now. We will just need to bring cqmain on our Arrow branch up-to-date with the main on upstream Arrow so that the parquet file writer changes are included.

[erezrokah]

This looks good to me. It seems most of the logic is implemented in the Arrow module.

As for the breaking changes, if those are easy to fix we can do them, otherwise we can make this a breaking change. Though I'm quite sure people will notice the change

[erezrokah]

We might want to add a . gitattributes file with the following content:

*.csv diff

I think that should allow these to be shown in the diff UI

[hermanschaaf]

I tried various ways but I think we can't force it--it's probably rendering it as binary because the files contain null bytes.

[erezrokah]

I think you need to change module github.com/cloudquery/filetypes to module github.com/cloudquery/filetypes/v2