fix: restore truncated identifiers and missing imports (NameError in process_document, prepare_dataset, plugin discovery)

[Anai-Guo]

What does this PR do?

Fixes a handful of identifiers that lost their last character, plus two missing stdlib imports. Each one makes the affected code path fail at runtime.

File	Problem	Effect
multimind/document_processing/document.py	text = documen / text = document.tex	DocumentProcessor.process_document always raises (NameError for str input, AttributeError for Document input), so process_file / process_text / process_url fail too
multimind/fine_tuning/peft_methods.py	return tokenized_datase	prepare_dataset raises NameError
multimind/fine_tuning/unified_peft.py (2×)	return tokenized_datase	same
multimind/vector_store/vector_store_enhanced.py	importlib.util used, never imported	the NameError is caught by the except, so every plugin in discover_plugins gets logged as a failed load
multimind/document_loader/document_loader.py	io.BytesIO used, never imported	loading a PDF from a URL raises NameError

All of these were flagged by pyflakes as undefined name. The repo's ruff config doesn't catch them because it doesn't enable F821.

Before / after

p = DocumentProcessor()
p.process_document("hello world. second sentence.")
p.process_document(Document(text="obj text here.", metadata={"k": 1}))

Before:

ERR NameError name 'documen' is not defined
ERR AttributeError 'Document' object has no attribute 'tex'

After:

OK [('hello world. second sentence.', {'chunk_index': 0, 'total_chunks': 1})]
OK [('obj text here.', {'k': 1, 'chunk_index': 0, 'total_chunks': 1})]

Checklist

• No breaking changes (7 lines added, 5 removed, no behaviour change beyond the fixes above)
• ruff check on the changed files: no new findings

🤖 Generated with Claude Code