Remove spurious 'single system message' warning from Anthropic provider

[timherby]

What this does

Removes a stale RubyLLM.logger.warn(...) call in Anthropic::Chat#build_system_content that fires on every request whenever a chat carries more than one :system-role message. The warning claims Anthropic only supports a single system message and that the gem will merge them -- neither is still true. Anthropic's system parameter has accepted an array of text content blocks (with per-block cache_control) since the prompt-caching launch in August 2024 (GA December 2024), and the gem already serializes correctly via flat_map over Media.format_content. Only the warning is stale.

At high request volume the warning floods logs for applications that intentionally split system content into a cached prefix + a per-conversation envelope. The Bedrock provider (lib/ruby_llm/providers/bedrock/chat.rb) handles the same case without warning, so this also aligns the two providers.

Adds direct unit coverage for build_system_content:

• returns [] for zero messages
• returns one block for one message
• returns two blocks (in order) for two messages
• does not log a warning when multiple :system messages are passed
• preserves cache_control on a Content::Raw block when paired with a plain-text block
• flattens mixed Content::Raw + plain-string content correctly

Fixes #767.

References:

• Anthropic Messages API: https://docs.anthropic.com/en/api/messages -- system: string | array of text content blocks
• Prompt caching with cache_control on system blocks: https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching
• anthropic-sdk-python ships system: Union[str, Iterable[TextBlockParam]]

Type of change

• Bug fix
• New feature
• Breaking change
• Documentation
• Performance improvement

Scope check

• I read the Contributing Guide
• This aligns with RubyLLM's focus on LLM communication
• This isn't application-specific logic that belongs in user code
• This benefits most users, not just my specific use case

Required for new features

Quality check

• I ran overcommit --install and all hooks pass
• I tested my changes thoroughly
  • For provider changes: Re-recorded VCR cassettes with bundle exec rake vcr:record[provider_name] (N/A -- this change is wire-format-neutral; no provider request/response shape changed, so existing cassettes still match)
  • All tests pass: bundle exec rspec spec/ruby_llm/providers/anthropic/chat_spec.rb (13 examples, 0 failures -- 6 new + 7 existing) and bundle exec rspec spec/ruby_llm/providers/anthropic_spec.rb (2 examples, 0 failures)
• I updated documentation if needed (no user-facing docs reference this warning)
• I didn't modify auto-generated files manually (models.json, aliases.json)

AI-generated code

• I used AI tools to help write this code
• I have reviewed and understand all generated code (required if above is checked)

API changes

• Breaking change
• New public methods/classes
• Changed method signatures
• No API changes

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.05%. Comparing base (ff39289) to head (9cb1ea9).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #768      +/-   ##
==========================================
+ Coverage   87.04%   87.05%   +0.01%     
==========================================
  Files         119      119              
  Lines        5596     5594       -2     
  Branches     1409     1407       -2     
==========================================
- Hits         4871     4870       -1     
+ Misses        725      724       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.