[codex] Implement local Iceberg accepted sink MVP

[malon64]

Refs #70

What Was Implemented

This PR implements the local filesystem MVP for the Iceberg accepted sink in Floe.

Core changes:

• Added sink.accepted.format: iceberg support to the accepted sink adapter registry and config validation.
• Implemented a local Iceberg writer adapter in floe-core that writes a valid Iceberg table layout under the configured local table root (metadata/, data/).
• Added append and overwrite handling through the existing sink.write_mode flow and accepted sink adapter architecture.
• Added scalar-type mapping from Floe/Polars output columns to a supported Iceberg scalar subset.
• Added report fields for accepted output metadata (write_mode, files_written, table_root_uri, snapshot_id when available) while keeping existing report fields backward-compatible.
• Updated docs/support matrix and Iceberg sink docs for the local MVP scope.

Behavior notes:

• append loads/registers the existing local Iceberg table metadata and commits a new snapshot.
• overwrite recreates table metadata at the same table root and writes new logical contents (no cleanup/GC of old files/metadata).
• Empty accepted dataframe overwrite/create initializes table metadata without writing data files or creating a snapshot.

Supported Scope / Explicit Non-Goals (Issue #70)

Supported in this PR:

• Accepted sink format: iceberg
• Local storage only
• write_mode: append, overwrite
• Scalar columns only (strings, booleans, signed ints, floats, date/time/datetime subset)
• Report metadata for Iceberg writes

Explicit non-goals in this PR:

• S3 / GCS / Glue catalogs
• Schema evolution
• Merge/upsert semantics
• Cleanup / garbage collection of orphaned files/metadata
• Advanced Iceberg table features beyond local unpartitioned MVP behavior

Test Coverage Summary

Added / updated tests:

• Local run() integration test that writes an Iceberg table on a temp dir and validates table layout + report fields.
• Unit tests for Iceberg writer append snapshot/version progression.
• Unit tests for overwrite logical replacement behavior.
• Unit tests for unsupported dtype errors (clear message).
• Unit tests for empty dataframe behavior.
• Config validation tests for local Iceberg acceptance and non-local Iceberg rejection.
• Updated existing report/CLI fixture tests for new accepted-output report fields.

Validation run in this branch:

• cargo fmt --all
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test --all

Follow-up Notes (#71 / #69 / #157)

• Iceberg sink: S3 implementation (object_store, no data catalog) #71 (S3 filesystem catalog) and Iceberg sink: GCS filesystem catalog #69 (GCS filesystem catalog) should reuse the Iceberg write preparation / type mapping / append flow introduced here, with object-store-backed file IO and catalog wiring.
• Iceberg sink with AWS Glue catalog (S3) — create/open/commit table #157 (AWS Glue catalog over S3) should build on the S3 Iceberg behavior and align with the future catalogs registry/config design.
• The overwrite path here intentionally prioritizes local MVP correctness and reviewability; cloud/catalog follow-ups can refine overwrite semantics once catalog implementations are in place.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9dc6d6a7a8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".