Implement support for duckdb_arrow_scan for ingesting data via Arrow

[phillipleblanc]

Adds support for calling the duckdb_arrow_scan when passed an FFI_ArrowArrayStream from the Arrow library that implements the https://arrow.apache.org/docs/format/CStreamInterface.html

This matches behavior that exists in the Go DuckDB library: https://github.com/marcboeker/go-duckdb/blob/c607539e645091c1e3f600f929229d8baa7166e4/arrow.go#L251

This also provides an alternative to the arrow-vtab module that is not zero-copy in all cases for Arrow -> DuckDB buffers, whereas the native C++ DuckDB code has better support for zero-copy Arrow -> DuckDB buffers.

[Maxxen]

Hello! Thanks a lot for the PR! Im about to cut v1.2.2 so in order to expedite getting this merged I forked this off into #489 with formatting/clippy fixes

[Maxxen]

Actually, ASAN reports memory leaks related to the test. I've done some digging and asked around the office from people with more arrow-knowledge, and my impression is that the arrow_scan_array method is most likely partly broken duckdb-side in that it has some weird assumptions regarding ownership and doesn't always free schemas properly. This PR seems related.

I guess these functions are deprecated for a reason. I don't know exactly when we will get around to replace them with something better, but we're looking into trying to fix what we think is causing this issue in particular - and if we do I'll check against this PR again.

[phillipleblanc]

Actually, ASAN reports memory leaks related to the test. I've done some digging and asked around the office from people with more arrow-knowledge, and my impression is that the arrow_scan_array method is most likely partly broken duckdb-side in that it has some weird assumptions regarding ownership and doesn't always free schemas properly. This PR seems related.

I guess these functions are deprecated for a reason. I don't know exactly when we will get around to replace them with something better, but we're looking into trying to fix what we think is causing this issue in particular - and if we do I'll check against this PR again.

Thanks for checking! I'll convert this PR to draft until the memory leak issue is fixed and/or a new C API is available that we can use.

[mlafeldt]

duckdb/duckdb#18246 lays the groundwork for a new Arrow C API.

[phillipleblanc]

Looks like the new C API will be coming out in DuckDB v1.4. I'll close this PR since we'll need a new one to integrate with the new C API.