[bug] Hub website: "Unable to parse markdown" — validator descriptions render with newlines stripped

Describe the bug

On the Guardrails Hub website, validator detail pages render their description with the message "Unable to parse markdown" followed by the raw description text with all line breaks collapsed into spaces.

Because the newlines are stripped, the markdown structure is destroyed: e.g. ## Description\n\nProvided an input dataset... becomes ## Description Provided an input dataset..., so the heading swallows the following paragraph and the markdown renderer falls back to dumping the unparsed raw text.

Example page: Arize Dataset Embeddings (hub://arize-ai/dataset_embeddings_guardrails) https://hub.guardrailsai.com/validator/arize-ai/dataset_embeddings_guardrails (the page shown below)

Important: the source content is correct

This is not a problem with the validator's README. The upstream source — Arize-ai/dataset-embeddings-guardrails/README.md — is well-formed markdown with proper line breaks and headings:

# Overview

| Developed by | Arize AI |
| --- | --- |
...

## Description

Provided an input dataset of "bad" prompts or LLM messages, this Guard will block...

### Benchmark results
...

The newlines are being lost somewhere in the Hub website's ingestion / rendering pipeline (the step that fetches the validator README and feeds it to the markdown component), not in the source repo. This means the bug likely affects many/all validator pages whose descriptions span multiple paragraphs, not just this one.

To Reproduce

1. Open the Hub validator page for "Arize Dataset Embeddings".
2. Observe the Description section renders Unable to parse markdown: ## Description Provided an input dataset of "bad" prompts... as a single run-on block instead of formatted markdown.

Expected behavior The description should render as formatted markdown (headings, paragraphs, lists, the benchmark table) exactly as it appears in the source README — with line breaks preserved.

Likely root cause / suggested fix The description string is being whitespace-collapsed (newlines → spaces) before it reaches the markdown renderer. Likely culprits: a \n-stripping normalization, an HTML-attribute/JSON round-trip that drops newlines, or a .replace(/\s+/g, ' ')-style sanitizer in the ingestion pipeline. The fix is to preserve \n characters in the description field end-to-end so the markdown parser receives the original structure.

Additional context This is a Hub website rendering issue (the website frontend source is not part of this repo), filed here as the canonical project tracker so it can be routed to the right team. A screenshot of the affected page is shown below.