Fix infinite retry on detected language rename

What does this MR do and why?

When Gitaly returns a new canonical name for a language whose language_id already exists under a different ProgrammingLanguage name, repository language detection repeatedly fails on the unique language_id constraint and retries forever. This can stall Sidekiq and, as reported, take instances down.

This MR:

• Resolves detected languages by language_id first, then by name.
• Updates existing records with the detected canonical name and language_id when it is safe.
• Returns the language_id match unchanged when the detected name is already taken by another record (no data repair).
• Matches resolved languages by language_id first when building repository language insertions.
• Recovers from ActiveRecord::RecordNotUnique by finding the conflicting record instead of retrying indefinitely.

This is the main fix, stacked on top of two refactoring MRs:

• Add detected language objects (!241475 - merged)
• Extract programming language resolver (!241485 - merged)

References

#602294 (closed)

How to set up and validate locally

1. Check out this branch (stacked on the resolver extraction branch):

   git switch 602294-fix-language-rename-infinite-retry

2. Run affected specs:

   bin/rspec spec/lib/gitlab/language_detection_spec.rb spec/lib/gitlab/language_detection/programming_language_resolver_spec.rb spec/services/projects/detect_repository_languages_service_spec.rb

3. Run RuboCop:

   bundle exec rubocop lib/gitlab/language_detection.rb lib/gitlab/language_detection/programming_language_resolver.rb spec/lib/gitlab/language_detection_spec.rb spec/lib/gitlab/language_detection/programming_language_resolver_spec.rb

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.