Update stale "Called for each generated token" docstring on rac_llm_component.h:199

Follow-up from #848 by @ayaangazali — thanks again for that PR!

What

core/include/rac/features/llm/rac_llm_component.h:199 still documents rac_llm_component_generate_stream's token_callback as "Called for each generated token" — the exact stale, one-token-per-call phrasing #848 just corrected on all four VLM streaming-callback declarations (core + Swift mirrors). The LLM decode loop already assembles UTF-8 internally (Utf8State), so this is a docs-only mismatch, not a live bug, but it leaves the public C ABI's own doc comment describing a contract the code doesn't follow.

Why it matters

Callers reasonably read the header doc as the contract. Leaving the LLM doc stale while VLM's was just fixed makes the split between the two confusing, and risks the same wrong "one call per token" assumption resurfacing in a future integration.

Suggested approach

Reword the doc comment the same way #848 reworded the VLM ones: describe delivery as UTF-8-safe text runs rather than one call per token.

Done when

• rac_llm_component.h's generate_stream doc no longer says "Called for each generated token".

Not blocking #848. @ayaangazali, you know this code well now — you're welcome to take this one if you're interested.

Opened with help from Claude Code and Codex.

Reviewed with help from Claude Code and Codex.