[BUG] FileSystemDocumentLoader#loadDocuments has a potential issue when using ApacheTikaDocumentParser

[vprudnikov]

Describe the bug
FileSystemDocumentLoader#loadDocuments has a potential issue when using ApacheTikaDocumentParser
As shown in the langchain4j-examples, this line of code doesn't work as expected:
List<Document> documents = loadDocuments(directoryPath, new ApacheTikaDocumentParser());

To Reproduce
Try to load at least 2 documents using the provided example.

Expected behavior
Each document is loaded independently.

Current behavior

• The first document contains its parsed content
• The second document contains the content of the first document plus its content
• The third document contains the content of the previous documents plus its content
• ... and so on

Please complete the following information:

• LangChain4j version: 0.30.0
• LLM(s) used: NA
• Java version: 21
• Spring Boot version (if applicable): NA

Additional context
It looks like the ApacheTikaDocumentParser is a stateful object. That's why it cannot be reused across all documents.
Probably, a better solution would be using a Supplier like so:
List<Document> documents = loadDocuments(directoryPath, () -> new ApacheTikaDocumentParser());
or
List<Document> documents = loadDocuments(directoryPath, filePath -> new ApacheTikaDocumentParser());
As a result, a new fresh instance of the parser will be provided each time.

[langchain4j]

@vprudnikov thank you so much for reporting!

[langchain4j]

Fixed by #1031