Claude Code GitHub Action needs visible PR feedback to replace Copilot

## Issue: Silent Analysis vs Visible PR Feedback

**Current State**:
The `anthropics/claude-code-action@v1` runs successfully but doesn't post visible review comments to PRs.

**Evidence from Testing**:
- ✅ Action executes perfectly (claude-sonnet-4-6, ~$0.95 cost, 6 turns)
- ✅ Comprehensive analysis (4+ minutes execution time)
- ✅ Uses specialized SRE prompts correctly
- ❌ **No visible output on PR** - defeats the purpose

**Expected Behavior (like GitHub Copilot)**:
- Post review comments directly on code lines
- Provide actionable feedback developers can see
- Enable PR conversation workflow
- Replace Copilot with Claude's superior analysis

**Current Logs Show**:
```
Running Claude Code via SDK (full output hidden for security)...
Rerun in debug mode or enable show_full_output: true
```

**Needed**:
- Parameter to enable PR comment posting
- Visible review output on code changes
- GitHub review integration (not just status updates)

**Use Case**: Enterprise teams want Claude to replace GitHub Copilot for PR reviews but need actual visible feedback in their workflow.

**Repository**: fullcast-io/SiteReliability
**PR for testing**: #151

Start with the anthropics/claude-code-action@v1 entry point and reproduce the behavior using the testing PR #151. Compare the current hidden SDK output and status updates with the requested line comments and review conversation; done means actionable feedback is visibly posted on the PR.

Written by the indexing model from the issue text.