docs: update DeepSeek pricing

[ai-bankofai]

Summary

• Align English and Chinese DeepSeek model, pricing, and promotion docs with the canonical bai-documents source.
• Update DeepSeek-V4-Flash standard pricing to 0.44 input, 0.0088 cache read, and 1.32 output.
• Update DeepSeek V4 Pro Busy pricing to 1.32 input, 0.0132 cache read, and 3.96 output.
• Clarify that V4 Pro currently uses Busy pricing, time-based pricing is not enabled, and V4 Flash remains free during the limited-time offer.
• Remove the stale GPT-5.5 price-ratio comparison and bump the documentation version from 1.3.21 to 1.3.22.

Validation

• yarn build (English and Chinese builds completed successfully; existing unrelated broken-link warnings remain).