ByteBufBsonDocument and RawBsonDocument simplifications

[rozza]

Rationale

ByteBufBsonDocument#clone used to return a RawBsonDocument. The recent changes returned a normal BsonDocument, which is potentially expensive depending on its usage.

The ByteBufBsonDocument changes also added complex iterator logic, when RawBsonDocument deferred to BsonDocument iterators. As iteration is essentially a hydrating mechanism, there is opportunity for improvements for both implementations. By changing the RawBsonDocument iterators to be more efficient, ByteBufBsonDocument can now utilize these efficiency gains by proxy, relying on the cachedDocument iterators.

This change both reduces the complexity of ByteBufBsonDocument and relies on an improved RawBsonDocument implementation.

Summary of changes

• ByteBufBsonDocument:

  • Simplify by returning RawBsonDocument from toBsonDocument, avoiding full BSON deserialization. When there are no sequence fields, the body bytes are cloned directly. When sequence fields exist, BsonBinaryWriter.pipe() merges the body with sequence arrays efficiently.
  • Use toBsonDocument for iterators. This eliminates the need for custom iterators (IteratorMode, CombinedIterator, createBodyIterator, and sequence iterators) since entrySet/values/keySet now delegate to the cached RawBsonDocument.

• RawBsonDocument:

  • Renamed toBaseBsonDocument to override the default toBsonDocument implementation.
  • Implemented the iterators so that they don't need to fully convert the document to a BsonDocument.

• Tests:

  • Updated ByteBufBsonDocumentTest iteration tests.
  • Updated ByteBufBsonArrayTest#fromValues as entrySet now returns RawBsonDocument instances.

JAVA-6010