Add OpenAI-compatible workers and live Lance/WebShart viewing

[bghira]

Summary

• add a BYOK OpenAI-compatible multimodal worker with endpoint pooling, adaptive concurrency, rate-limit handling, retries, and environment-only provider credentials
• support backend-neutral inference configuration while preserving existing vLLM configuration compatibility
• add direct HTTP range loading for remote WebShart samples to avoid downloading complete tar shards
• keep contributor totals and monitor activity accurate during live orchestration
• update the dataset viewer to read live Lance stores, resolve WebShart image previews by byte range, handle missing values, and fit images within the preview viewport
• document the API worker workflow and include an example configuration

Testing

• pytest -q --disable-warnings
• 467 passed

[codecov]

Codecov Report

❌ Patch coverage is 88.88889% with 84 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/caption_flow/workers/openai_compatible.py	90.30%	48 Missing ⚠️
src/caption_flow/viewer.py	83.16%	17 Missing ⚠️
src/caption_flow/orchestrator.py	56.25%	7 Missing ⚠️
src/caption_flow/cli.py	70.00%	6 Missing ⚠️
src/caption_flow/storage/manager.py	93.75%	3 Missing ⚠️
src/caption_flow/workers/caption.py	50.00%	2 Missing ⚠️
src/caption_flow/processors/webdataset.py	98.61%	1 Missing ⚠️

Files with missing lines	Coverage Δ
src/caption_flow/processors/webdataset.py	90.02% <98.61%> (+1.48%)	⬆️
src/caption_flow/workers/caption.py	63.52% <50.00%> (+0.20%)	⬆️
src/caption_flow/storage/manager.py	73.76% <93.75%> (+1.58%)	⬆️
src/caption_flow/cli.py	47.16% <70.00%> (+0.41%)	⬆️
src/caption_flow/orchestrator.py	41.74% <56.25%> (+1.87%)	⬆️
src/caption_flow/viewer.py	53.33% <83.16%> (+25.41%)	⬆️
src/caption_flow/workers/openai_compatible.py	90.30% <90.30%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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