[FIX] web: Fix download.js handling of large content.

When handling a non-200 reponse with a large payload (>65536 bytes),
download.js fails to decode the payload properly on WebKit.

This is because the content is parsed using WebKit's
`DOMParser.parseFromString`, which creates several Text nodes if the
text would exceed 65536 bytes. Then, only the textContent of the second
Text node is passed to `JSON.parse()`, which fails because it is not
valid JSON.

See https://stackoverflow.com/questions/67738121/in-what-cases-do-browsers-create-multiple-adjacent-text-nodes/67774415#67774415
and https://github.com/WebKit/WebKit/blob/68ae0fde5f959e056fbd6700f1ca7fa652cd1ffa/Source/WebCore/html/parser/HTMLConstructionSite.cpp#L584-L592

closes odoo#160815

Taskid: 3790302
X-original-commit: a68348f
Signed-off-by: Luca Vitali (luvi) <[email protected]>

addons/web/static/src/core/network/download.js

@@ -542,7 +542,7 @@ export function configureBlobDownloadXHR(
                 const contents = decoder.result;
                 const doc = new DOMParser().parseFromString(contents, "text/html");
                 const nodes =
-                    doc.body.children.length === 0 ? doc.body.childNodes : doc.body.children;
+                    doc.body.children.length === 0 ? [doc.body] : doc.body.children;
 
                 let error;
                 try {