The new Vector 2022 skin is now using the new Codex Typeahead search. However, by default it uses the MediaWiki search API. Whereas on Wikidata, we need it to use wbsearchentities in order to actually search through Items, Properties, and Lexemes. Currently, the new Vector skin still contains some code to maintain [backwards compatibility with the Wikidata hack](https://gerrit.wikimedia.org/r/825258).
To make the new Vecor search use wbsearchentities, we need to create a custom search client and store it at [wgVectorSearchClient](https://github.com/wikimedia/Vector/blob/2321eeb0547e485504775d1118089377a43f1828/resources/skins.vector.search/restSearchClient.js#L83-L85) in order to override the default implementation. The basic feasibility is outlined by this [Proof of Concept](https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/817312).
**Aceptance Criteria**:
* [ ] The new vector search uses wbsearchentities
* [ ] The placeholder for the article images are not shown (that might need some config change or something)
* [ ] Selecting results with the Keyboard works as expected
**Notes**:
* The second aspect that's missing for full feature parity with the current solution is the "load more" behavior that we have with the current Wikidata search. That is currently waiting on implementation in {T306933}, stalled on {T306932}, which is intended to be implemented in context of {T313477}.
** whether we can switch to the new search after the current story is done, but before that "load-more" behavior is implemented, that is a product decision, and should probably be made after this story has been implemented