Chewy.close_client closes Elasticsearch connections

[AlfonsoUceda]

Chewy.client is memoized per thread, so every thread that touches Chewy builds its own client holding its own connections to Elasticsearch. Neither elasticsearch-ruby nor elastic-transport expose a way to close those connections, so they only get released when the GC reclaims the dead thread's client. In long-lived multi-threaded processes that churn threads — Sidekiq replaces a thread on job failure — dead-but-not-yet-collected clients pile up and leak file descriptors. See #951.

This adds Chewy.close_client, which closes the current thread's connections (via Faraday, guarded by respond_to? so it's adapter-agnostic) and drops the thread-local so the next Chewy.client rebuilds fresh. Users wire it into a Sidekiq death handler to release connections immediately instead of waiting for GC. Note the default net_http adapter opens/closes per request and doesn't leak; this matters for persistent adapters (httpclient, typhoeus, patron).

Testing

bundle exec rspec spec/chewy_spec.rb -e '.close_client' spec/chewy/elastic_client_spec.rb -e '#close' — all green. Specs use doubles, no live cluster needed.

---

Before submitting the PR make sure the following are checked:

• The PR relates to only one subject with a clear title and description in grammatically correct, complete sentences.
• Wrote good commit messages.
• Commit message starts with [Fix #issue-number] (if the related issue exists).
• Feature branch is up-to-date with master (if not - rebase it).
• Squashed related commits together.
• Added tests.
• Added an entry to the changelog if the new code introduces user-observable changes. See changelog entry format for details.