fix(azure): correct GPT-5.6 Luna pricing

[vankridee]

TL;DR

• Correct Azure GPT-5.6 Luna input, output, and cache pricing.
• Add regression coverage for both Azure Luna model entries.

Changes

• Updated azure/gpt-5.6-luna pricing to match the expected Azure rates.
• Updated azure/us/gpt-5.6-luna pricing to match the expected Azure rates.
• Added regression tests covering both model pricing entries.

Testing

• uv run pytest tests/test_litellm/test_model_prices_schema.py -v
• Result: 20 passed

Related Issue

Fixes #36094

[CLAassistant]

All committers have signed the CLA.

[greptile-apps]

Greptile Summary

Corrects input, output, cache-read, long-context, and priority pricing for the global and US Azure GPT-5.6 Luna entries

• Adds exact-value regression coverage for both modified canonical price entries
• Does not synchronize the bundled fallback price map used by local or failed-remote loading

Confidence Score: 4/5

The fallback pricing map must be synchronized before merging, or supported local and offline deployments will continue using incorrect Luna rates

The canonical rates and focused assertions agree, but runtime can load an unchanged bundled backup that preserves the pricing defect

Files Needing Attention: model_prices_and_context_window.json and litellm/model_prices_and_context_window_backup.json

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Corrects both canonical Azure Luna entries, but leaves the supported bundled fallback map with obsolete rates
tests/test_litellm/test_model_prices_schema.py	Covers every pricing field changed in both canonical entries, but does not exercise fallback-map synchronization

_{Reviews (1): Last reviewed commit: "fix(azure): correct GPT-5.6 Luna pricing" | Re-trigger Greptile}