send initialCanonicalUrl in array format to fix crawler issues (#69370)
The initial RSC payload currently provides the `initialCanonicalUrl`
value to initialize the router state during the SSR render. However as
noted in #53274, it seems crawlers like Googlebot will crawl anything
URL-like that it finds in the document, regardless of whether or not
it's actually rendered as a link somewhere (which this isn't).
This splits the `initialCanonicalUrl` value into parts and then joins it
when it's ready to be consumed. That way when it's embedded in the
document, it's not in a path-like format.
Fixes #53274
Closes NDX-243