This would be useful:

1. To add more bootstrap validation in a future version (6, probably), so that nobody tries to sort on tokenized fields
2. To integrate the Elasticsearch concept of normalizer ( HSEARCH-2659 In Progress )

Regarding the API, we'd probably just add @AnalyzerDef, @Analyzer + programmatic API equivalent + AnalyzerProvider equivalent. Maybe rename LuceneAnalyzerDefinitionProvider to LuceneAnalysisDefinitionProvider, just to be consistent?