[Bot] Update Inference Providers documentation

[HuggingFaceInfra]

This PR automatically upgrades the @huggingface/tasks and @huggingface/inference packages and regenerates the Inference Providers documentation by running:

cd scripts/inference-providers
pnpm update @huggingface/tasks@latest @huggingface/inference@latest
pnpm run generate

This PR was automatically created (and merged) by the Update Inference Providers Documentation workflow.

---

Note

Low Risk
Auto-generated documentation and dependency pin only; no runtime or application logic changes.

Overview
Bumps @huggingface/tasks from 0.21.2 to 0.21.4 in scripts/inference-providers (lockfile included) and regenerates Inference Providers markdown so InferenceSnippet providersMapping values match current provider catalogs.

Example model updates: fal-ai image-to-image snippets now use fal/Qwen-Image-Edit-2511-Multiple-Angles-LoRA (was Alissonerdx/BFS-Best-Face-Swap). HF Inference token-classification examples use openai/privacy-filter instead of dslim/bert-base-NER.

Provider capability shifts in docs: SambaNova gains a Chat Completion (VLM) section and appears in shared VLM snippet maps (meta-llama/Llama-4-Maverick-17B-128E-Instruct). Hyperbolic loses its provider-level VLM section and is removed from the aggregated conversational VLM mappings on chat-completion and image-text-to-text task pages.

^{Reviewed by Cursor Bugbot for commit 4c055eb. Bugbot is set up for automated code reviews on this repo. Configure here.}