fix(eval): thread-safe progress callbacks for --concurrency

[macekond]

Summary

• Wrap Rich Console.print() calls in _make_progress_callbacks with threading.Lock
• Fixes deadlock when using --concurrency 2+ with piped stdout (e.g. background processes)
• Add test_progress_callbacks_thread_safe test

Root Cause

_make_progress_callbacks creates closures that call console.print(). When run_items dispatches to ThreadPoolExecutor, multiple threads invoke these callbacks simultaneously. Rich Console is not safe for concurrent writes to piped stdout/stderr, causing a deadlock where the process hangs at 0% CPU with no output.

The lock serializes only the console.print() calls — actual evaluation work (AI chat, scoring) remains fully concurrent.

Reproduction

# This hangs with concurrency on large datasets:
gd-eval run --concurrency 4 --dataset large_dataset/ --model gpt-5.2 &

# Workaround before this fix:
gd-eval run --concurrency 4 --quiet --dataset large_dataset/ --model gpt-5.2 &

Test plan

• test_progress_callbacks_thread_safe — 8 concurrent threads calling all 3 callbacks 50 times
• Verified --concurrency 4 works on 10-item dataset with progress output
• Verified --quiet mode still works (callbacks are None, no lock needed)
• Verified sequential mode (concurrency=1) unaffected

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.21%. Comparing base (04bfbc5) to head (b526366).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1671   +/-   ##
=======================================
  Coverage   79.21%   79.21%           
=======================================
  Files         232      232           
  Lines       15809    15809           
=======================================
  Hits        12523    12523           
  Misses       3286     3286           

☔ View full report in Codecov by Harness.
📢 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.