Remove usage of IndexSearcher#search(Query, Collector) from join package

[msfroh]

Description

Relates to https://github.com/apache/lucene/issues/12892

For global ordinal-based join, we can support concurrent search. For numeric and term-based joins, we fail if we're called from a multithreaded searcher.

I can implement concurrent versions of the other join collectors, but wanted to get a first pass that removes the uses of IndexSearcher#search(Query, Collector).

Note that for the cases that still assume a single-threaded searcher, I used a version of CollectorManager#wrap(Collector) from https://github.com/apache/lucene/pull/13735, with my guess for where it will end up based on feedback so far.