[Aryex]

Summary

Fixes FT Search builder options that silently dropped what the caller asked for, or crashed on a "not given" value. Behaviour now matches the other GLIDE clients (Python, Java, Node, Go). All changes are in Ruby-side request construction, before the command reaches GLIDE Core.

Also removes the "redis-rb conventions" rule from AGENTS.md: GLIDE clients are the reference for command behaviour.

Issue link

• Closes Search Module: Preserve explicit empty FT.CREATE stopword lists
• Closes Search Module: Reject sort_order when sort_by is omitted
• Related Search Module: Leaking NoMethodError (fixes the clauses: nil case)

Changes

• CreateOptions stopwords: (Search Module: Preserve explicit empty FT.CREATE stopword lists #386): [] used to be dropped, so the server used its default stopwords. It now sends STOPWORDS 0, which turns stopword removal off. nil still sends nothing.
• AggregateOptions clauses: (Search Module: Leaking NoMethodError #387): the default is now nil, and nil means no clauses. Before, clauses: nil raised NoMethodError.
• SearchOptions sort_order: (Search Module: Reject sort_order when sort_by is omitted #388): the default is now nil.
  • An explicit sort_order without sort_by raises ArgumentError, "sort_order requires sort_by". Before, it was silently ignored.
  • sort_by with no order now sends only SORTBY <field> and lets the server apply its default (ascending). Before, Ruby always added ASC. Results are unchanged.
• AGENTS.md: removed the "Prefer matching redis-rb method signatures and return types" rule.

Behaviour	Ruby (this PR)	Python	Java	Node	Go
stopwords: []	STOPWORDS 0	STOPWORDS 0	STOPWORDS 0	STOPWORDS 0	dropped
clauses: nil	no clauses	no clauses	no clauses	no clauses	no clauses
order without sort field	raises	raises	raises	raises	raises
sort field, no order	SORTBY f	SORTBY f	SORTBY f	SORTBY f	SORTBY f

Limitations

• Behaviour change: code that passes sort_order: without sort_by: used to be silently ignored and now raises.
• No runtime type checks: wrong argument types, such as return_fields: "title" from Search Module: Leaking NoMethodError #387, aren't checked at runtime. Ruby relies on the documented YARD types, and none of the other clients validates these at runtime either.
• No CHANGELOG entry: FT Search is new in this unreleased version.

Testing

• Added unit tests:
  • test/unit/commands/search_builder_test.rb: test_create_options_stopwords_empty, test_create_options_rejects_empty_stopwords_with_no_stopwords, test_ft_create_builder_with_empty_stopwords
  • test/unit/commands/search_query_test.rb: test_sort_order_asc, test_sort_order_nil, test_sort_order_requires_sort_by, test_ft_search_sort_order_requires_sort_by (checks that nothing is sent), test_return_fields_empty, test_params_empty
  • test/unit/commands/search_aggregate_test.rb: test_aggregate_options_nil_clauses
• Updated tests: existing tests now expect SORTBY <field> without ASC when no order is given. test_sort_by_default_asc_and_desc was renamed to test_sort_by_without_order_and_desc.
• Removed: test_create_options_empty_stopwords_emits_nothing, which asserted the Search Module: Preserve explicit empty FT.CREATE stopword lists #386 bug.
• Local run on Ruby 3.4: bundle exec rake test:unit (524 tests, 0 failures) and bundle exec rubocop (no offenses). Integration tests run in CI.

Checklist

Before submitting the PR make sure the following are checked:

• This Pull Request is related to an issue.
• Commit message describe your changes
• Commits are signed off (git commit -s) per the DCO.
• Tests are added or updated.
• CHANGELOG.md and documentation files are updated. (No CHANGELOG entry: FT Search is new in this unreleased version.)
• Linters have been run (bundle exec rubocop) and pass.
• Destination branch is correct - main.

[jamesx-improving]

LGTM