Review and replace comparison slider images

[planetf1]

Problem

The before/after slider images don't align with the project's positioning strategy:

• without-mellea.png shows a raw system prompt (prose) — not competing code
• with-mellea.png shows a 30+ line Granite Guardian + RAG pipeline using IBM_GRANITE_GUARDIAN_3_0_2B — this is Layer 4 (specialist/Granite-specific), not the gateway experience

The positioning research recommends leading with @generative (Layer 1) and de-emphasizing Granite-specific features in adoption content.

Suggested replacement

Show @generative vs LangChain OutputParser — the strongest demo per the research doc:

Side	Content
Without Mellea	LangChain PydanticOutputParser + prompt template + chain + try/except (~40-50 lines)
With Mellea	@generative decorator with Pydantic return type (~8 lines)

This targets the primary audience (frustrated LangChain users) and showcases the genuine unique differentiator.

Questions to resolve

1. Should we label the sides explicitly (e.g. "LangChain" / "Mellea") or keep the generic "Without/With" framing?
2. Should these be screenshots of an editor, or rendered code blocks?
3. Is there a specific @generative example from the codebase to use as the canonical "hero" demo?