fix(deep-research): keep oversized context from collapsing to empty

[XiaojuCH]

Fixes #1770.

This PR prevents deep-research context trimming from collapsing to an empty string when the first retained context item alone exceeds MAX_CONTEXT_WORDS.

What changed:

• update trim_context_to_word_limit() in gpt_researcher/skills/deep_research.py
• when the first candidate item already exceeds the cap and nothing has been retained yet, keep a truncated version of that item instead of returning an empty context
• add focused regression tests in tests/test_deep_research_parsing.py

Why:

• deep research currently joins final_context into self.researcher.context
• if trim_context_to_word_limit() returns [], the writer receives "" for {context}
• this matches the failure mode described in Deep mode: writer receives empty {context} placeholder; report rejection across all params (post-PR-#1637, current master) #1770, where deep research completes but the writer rejects the report because no research material is present

Validation:

• .\.venv\Scripts\python.exe -m pytest tests/test_deep_research_parsing.py -q
• 25 passed