Better Iceberg support

[scsmithr]

Initial iceberg support was added in #1382. This issue serves to track further improvements to our iceberg support.

Grouped roughly by priority. Non-exhaustive, and priorities subject to change as we get feedback.

In addition to these, we'll want to look at upstreaming stuff to iceberg-rust.

High

• Understand where we fall apart when it comes to reading V1 format.
• Add tests for GCS/S3
• Add tests with iceberg table containing all types supported by iceberg

Medium

• Support handling row-level deletes
• Support schema evolution
• Default value handling
• Properly provide partitions to ParquetExec (perf)

Low

• Surface statistics to ParquetExec (perf)

Unknown

• CREATE EXTERNAL TABLE support (catalog?)
• Write support
• ORC, Avro support

[vrongmeal]

I've got a v1 dataset working but we need some old datasets in order to fix more stuff. Moreover, currently I only have serde(default) to get it working, but in longer term we would want to have different structs for v1 and v2.