Algorithm in BlobsExtractor works wrong for certain documents

We have pretty complex document types in our system.
And sometimes an exception happens on attempt to get all blobs from a document of such type.

The problem is in behavior of BlobsExtractor class.
It tries to find all blob properties in a document metadata.
But it fails in case when a complex property with a blob field goes after a complex property without blob.

If you apply changes (attached) to the existing test data of nuxeo-core-api project, TestBlobsExtractor.test will fail.

Provided by Gagnavarslan.