Propagate UTF-8 text-run streaming-docstring wording to VLM headers

Follow-up from #928 by @Sarah-Mahmoud-29 -- thanks again for that PR!

What

#928 corrected the LLM streaming callback docstring at core/include/rac/features/llm/rac_llm_component.h:199 from "Called for each generated token" to "Called for each UTF-8-safe text run generated by the model." The identical stale wording -- "Called for each generated token" -- is still in four sibling locations that go through the same rac::tokens::StreamFilter chunking path: core/include/rac/features/vlm/rac_vlm_component.h:177, core/include/rac/features/vlm/rac_vlm_types.h:409 (and :453), core/include/rac/features/llm/rac_llm_types.h:265, and the mirrored (real file, not a symlink) Swift-bindings copy bindings/swift/Sources/RunAnywhere/CRACommons/include/rac_vlm_component.h:166 (plus rac_vlm_types.h:373).

Why it matters

Callers implementing the VLM/LLM streaming callbacks read one doc that promises per-token delivery and another that promises per-run delivery for the same underlying chunking behavior across sibling APIs, which is confusing and can lead to incorrect buffering assumptions on the caller side.

Suggested approach

Apply the same wording correction at all sites in one pass, keeping the Swift-bindings mirror in sync since it is a real duplicated file rather than a symlink.

Done when

• All VLM header sites (core + Swift-bindings mirror) describe streaming delivery consistently with rac_llm_component.h:199.

Not blocking #928. @Sarah-Mahmoud-29, 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.