feat(python): Python binding for iceberg-rust Schema

[abnobdoss]

Status

Blocked for now while the Predicate binding stack settles. This draft is runtime-only; Python typing stubs are deferred to a separate package-wide follow-up PR.

Summary

Adds a Python binding for iceberg::spec::Schema as an opaque pyiceberg_core.schema.Schema handle constructed from Iceberg schema JSON.

• Schema.from_json(s) parses V1 or V2 schema JSON and lets iceberg-rust serde enforce schema validity
• schema_id(), highest_field_id(), column_names(), identifier_field_ids() expose cheap schema metadata; identifier field IDs are returned in ascending order
• find_field_by_name(name) does case-sensitive dotted-path lookup and returns {id, name, type, required} or None
• field_by_id(id) returns the same field dict shape and raises KeyError when absent
• to_json() emits parseable schema JSON for semantic round trips
• to_arrow_schema() exports a pyarrow.Schema; field IDs are preserved in PARQUET:field_id metadata
• __arrow_c_schema__() implements the Arrow PyCapsule Interface with capsule name "arrow_schema"
• _capsule() returns a PyCapsule named "iceberg_core_schema" wrapping Arc<Schema> for future sibling modules in this binding crate

The field dict keeps type as the Iceberg spec JSON representation rather than exposing a parallel Python type tree. That keeps this PR focused on an opaque schema handle while preserving enough information for callers that need to inspect a field type.

Files changed

• bindings/python/src/schema.rs - Schema binding implementation
• bindings/python/src/lib.rs - registers the schema submodule
• bindings/python/Cargo.toml - adds the explicit serde_json dependency used by from_json() and to_json()
• bindings/python/tests/test_schema.py - schema binding tests

Verification

• maturin build --release --out dist - clean, zero warnings
• pytest bindings/python/tests/test_schema.py - 32 passed
• cargo test -p iceberg --lib - 1294 passed

Design notes

Arc<Schema> makes clones cheap and gives _capsule() a clean ownership story: each capsule owns its own Arc clone, so the capsule remains valid after the Python Schema object is dropped.

#[pyclass(..., from_py_object)] is included so follow-up methods in this binding crate can accept Schema as a typed Python argument under PyO3 0.28 without relying on deprecated implicit extraction behavior.

Typing stubs and py.typed are intentionally deferred to a package-wide typing PR so this feature PR only changes runtime behavior.