Clarify the status of Kuromoji dictionaries

### Description

While refactoring the gradle code/data generation code I stumbled across the fact that we currently have two different tasks for generating the same set of output files - one is compileMecab, the other is compileNaist. They use different inputs but write to the same output files.

There is also this patch, which seems to be hanging or abandoned -
https://github.com/apache/lucene/pull/12517/files

I don't have any experience with Kuromoji... is there any reason to keep both inputs? Should it be configurable at runtime somehow?

At the moment, to get naist dictionary, you need to generate it by hand and recompile Lucene.

Start by reading the Gradle code/data-generation tasks compileMecab and compileNaist, comparing their inputs and shared output files. Review the status and intent of pull request 12517, then determine whether both dictionary inputs should remain or be configurable at runtime. Done means the supported Kuromoji dictionary workflow and the required code or documentation changes are clearly decided.

Written by the indexing model from the issue text.