feat: Add Snapshot::into_builder for updating existing snapshots

[alexanderbianchi]

What changes are included in this PR?

Add Snapshot::into_builder() to modify an existing snapshot without manually copying every unchanged field:

let updated = snapshot.into_builder().with_summary(summary).build();

TableMetadata already exposes the same conversion:

let builder = metadata.into_builder(None);

That delegates to TableMetadataBuilder::new_from_metadata(metadata, None). Snapshots currently have only a builder for constructing a new value, so even replacing a summary requires repeating IDs, sequence number, timestamp, manifest location, and other fields.

into_builder() returns a SnapshotUpdateBuilder that owns the existing snapshot and replaces only explicitly set fields. The existing Snapshot::builder() and its compile-time required-field checks are unchanged: its typed builder does not permit overwriting populated fields, so it cannot serve as an editing builder directly. No dependencies or serialization changes.

Motivation

Use case: DataFusion Distributed #700, where fixture tests replace snapshot summaries to exercise missing statistics. This review comment highlights the resulting snapshot-reconstruction boilerplate.

This constructs a snapshot value only; table metadata updates and their snapshot-ID/sequence-number validation remain unchanged.

Are these changes tested?

• Unit test covers an unchanged round trip, replacing only the summary while preserving all other fields (including encryption and row lineage), field overrides, and clearing optional IDs.
• cargo test -p iceberg --lib --locked — 1,675 passed.
• cargo test -p iceberg --doc into_builder --locked — passed.
• cargo clippy -p iceberg --lib --tests --all-features --locked -- -D warnings — passed.
• Public API snapshot regenerated with cargo public-api -p iceberg --all-features -ss.
• cargo fmt --all -- --check and git diff --check — passed.

AI Disclosure

Implementation, tests, and this description were prepared with AI coding assistance.