Preserve JSON string upload URLs

When AtlasCloud's media upload responds with a valid JSON string containing a URL, such as "https://...", this branch now wraps the parsed string in { error: { message: text } } instead of returning the string. uploadAtlasCloudMedia immediately passes this value to extractAtlasCloudUploadedMediaUrl, which has an explicit top-level string URL case, so those uploads regress from succeeding to AtlasCloud media upload did not return a usable URL; the null crash guard should not discard string primitives that existing extraction logic can handle.

Useful? React with 👍 / 👎.

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Exclude arrays in the parsed-body guard to preserve response contract.

On Line 2719, the object check still accepts arrays (typeof [] === "object"), so array JSON payloads bypass the fallback and are returned as Record<string, unknown>. That contradicts the intended behavior for non-object payloads and can leak invalid body shapes downstream.

-			body =
-				typeof parsed === "object" && parsed !== null
+			body =
+				typeof parsed === "object" &&
+				parsed !== null &&
+				!Array.isArray(parsed)
 					? (parsed as Record<string, unknown>)
 					: {
 							error: {
 								message: text,
 							},
 						};

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@apps/gateway/src/videos/videos.ts` around lines 2719 - 2721, The object type
guard in the parsed-body condition is accepting arrays because typeof array
returns "object" in JavaScript, which causes array payloads to bypass the
fallback handler and be returned as Record<string, unknown>, violating the
intended response contract. Modify the guard expression that checks `typeof
parsed === "object" && parsed !== null` to additionally exclude arrays by adding
an Array.isArray check, ensuring only plain objects pass through while arrays
fall through to the fallback case.