fix: HTMLToDocument generates noisy ERROR logs when receiving empty ByteStream objects

[deep-rloebbert]

Describe the bug
When HTMLToDocument receives a ByteStream with data=b"" (empty), it calls trafilatura.extract() on the empty bytes. This triggers lxml to emit several ERROR/WARNING level log events that surface as noise in production logs, even though the component already has a try/except that gracefully skips the source.

Error message

  lxml parsing failed: Document is empty                                                                                                    
  lxml parser bytestring Document is empty                                                                                                  
  empty HTML tree: None                                                                                                                     
  discarding data: None                                                                                                                     

The errors appear in pairs (lxml fires them during multiple internal parse attempts), and the surrounding pipeline continues normally, but the log output looks like a serious failure.

Expected behavior
HTMLToDocument should silently skip (or log at DEBUG/WARNING level) any ByteStream whose data is empty, without triggering lxml at all.

To Reproduce

from haystack.components.converters.html import HTMLToDocument                                                                            
from haystack.dataclasses import ByteStream                                                                                               
                                                                                                                                            
converter = HTMLToDocument()                                                                                                              
# Simulate what LinkContentFetcher emits on failure/redirect with raise_on_failure=False                                                  
empty_stream = ByteStream(data=b"")                                                                                                       
empty_stream.mime_type = "text/html"                                                                                                      
                                                                                                                                            
result = converter.run(sources=[empty_stream])                                                                                            
# Four lxml ERROR/WARNING log lines are emitted even though result == {"documents": []}                                                   

FAQ Check

• Have you had a look at our new FAQ page?

System:

• Haystack version: haystack-ai 2.28.0

[Srivatsa03]

Hi! I'd like to work on this. Plan: short-circuit empty ByteStream objects in HTMLToDocument.run() before calling trafilatura, so lxml never emits those parse errors.