Only fetch metadata for records used in constructing normalized concept

Eg, many Drugs@FDA records include xrefs to RxNorm entities that we don't load. Should the normalize/ result for that record include RxNorm meta (probably not)?

No files, tests, or entry points are named. Start by tracing how normalization results fetch and attach metadata for Drugs@FDA records, then compare that behavior with unloaded RxNorm entities referenced by xrefs. Done means the expected inclusion or exclusion of RxNorm metadata is decided and covered by an appropriate regression test.

Written by the indexing model from the issue text.