web: fix refered object extractions when page is in context of web site

Base_extractReferredObjectDictFromHtml except that base url to have a
leading /, because we ignore the first character of the normalized URL
returned by Base_normalizeUrlPathname.

By using / as base URL when we are in the context of the website, the
paths passed to Base_normalizeUrlPathname always start by / so we can
safely drop / from the normalized path returned by
Base_normalizeUrlPathname
3 jobs for fix/web-page-implicit-successor-getDocumentValue
in 0 seconds, using 0 compute credits, and was queued for 0 seconds