docs: document which query_by field controls query tokenization

[tharropoulos]

Change Summary

• add a :::warning callout under the Query parameters table in api/search.md explaining the search query is tokenized once using the highest-weighted query_by field's locale, token_separators, symbols_to_index and stemming
• note in the callout that those tokens are reused for every queried field while each field is still indexed, faceted and highlighted with its own settings
• append a short pointer to the query_by cell linking down to the callout
• clarify that the highest-weighted field is the first listed in query_by unless query_by_weights reorders precedence
• add pointers on the token_separators, symbols_to_index, locale and stem settings in api/collections.md marking them index-time only while query parsing uses the top query_by field
• add a matching callout to the guide/locale.md multilingual guide covering how the top query_by field decides tokenization when fields use different locales
• replicate the api/search.md and api/collections.md notes across versions 28.0 through 30.2

PR Checklist

• I have read and signed the Contributor License Agreement.