Improve memory usage when adding attachments to imported pages

The objective of this ticket is to improve the batch import API to use java InputStream API when adding attachments to imported pages, the improvement is useful to avoid import failure caused by memory issues when attaching big files which are currently loaded entirely on memory.

Steps to reproduce

• On an XWiki instance (version 14.4.6) install the Batch Import application version 2.3
• Create a CSV file containing the data of the page to be imported, see screenshot
  
• Create a zip file (files.zip) that contains a big file, (file size > 1GB), named attachment.zip
• Create a test page and edit it using object editor
• Attach an object of class to " BatchImport.BatchImportClass" to the test page
• Set the object properties as follows :
  • fileupload = true
  • datafilename = The path to the zip file that contains the attachments (path to files.zip)
• Save the changes
• Open the test page in view mode and follow the batch import steps until the last one (import), here is the mapping to be used :
  

Expected result : The page has been correctly imported and the file attached to the page

Current result : The import failed with a "java.lang.OutOfMemoryError: Java heap space" error, see screenshot :