Cache NodeIdentifier lookup in QueryBuilderState.bindMethod()

When we are building a query, we end up creating a NodeIdentifier for the found leaf. Finding the leaf involves quite a few string operations, hence we want to side-step it. Furthermore, once we have located the node we follow up with a caching NodeIdentifier.create(), which could also be cached.

Add a cache to DefaultQueryBuilder, so that all queries with particular code end up reusing a common cache of these.