refactor(gravsearch): Simplify filtering of incoming links

[benjamingeer]

In #1626 (comment), we found that we rely on MainQueryResultProcessor.getRequestedValuesFromResultsWithFullGraphPattern to prevent extra incoming links in a query like this:

PREFIX incunabula: <http://0.0.0.0:3333/ontology/0803/incunabula/simple/v2#>
PREFIX knora-api: <http://api.knora.org/ontology/knora-api/simple/v2#>

CONSTRUCT {
    ?region knora-api:isMainResource true .
    ?region knora-api:isRegionOf ?page .
    ?page knora-api:isPartOf ?book .
    ?book incunabula:title ?title .
} WHERE {
    ?region a knora-api:Region .
    ?region knora-api:isRegionOf ?page .
    ?page a incunabula:page .
    ?page knora-api:isPartOf ?book .
    ?book a incunabula:book .
    ?book incunabula:title ?title .
    FILTER(?title = "Zeitglöcklein des Lebens und Leidens Christi")
}

Here we get four regions on different pages in the same book. Without MainQueryResultProcessor.getRequestedValuesFromResultsWithFullGraphPattern, all the links from all pages to the book show up as incoming links each time the book appears as a dependent resource in the response to the API request. This is because ConstructResponseUtilV2.splitMainResourcesAndValueRdfData doesn't consider the response tree structure when it starts constructing link values for incoming links. It would be better to handle this logic in ConstructResponseUtilV2.splitMainResourcesAndValueRdfData; then we could get rid of MainQueryResultProcessor.