Use canonical DeepSeek model id (~deepseek/deepseek-v4-flash-latest) and update pricing/README

[astrovm]

Motivation

• Standardize the DeepSeek model identifier to the canonical gateway-aware id and align documentation and pricing so billing/token limits match the provider naming.
• Ensure code paths that special-case DeepSeek (token limits, pricing lookups, defaults) work with the updated model id.

Description

• Replaced occurrences of deepseek/deepseek-v4-flash with ~deepseek/deepseek-v4-flash-latest across code, tests and README.md.
• Updated MODEL_PRICING_USD_MICROS to use the new model key and adjusted lookups in api/ai/pricing.py and estimate_chat_reserve_credits to reference the canonical id.
• Adjusted the chat_output_token_limit check, default model constants PRIMARY_CHAT_MODEL and SUMMARY_MODEL in api/index.py, and benchmark entries in benchmark_bot.py to use the new id and updated context/pricing text in the README.
• Updated tests and fixtures to expect the new model identifier in billing, provider runtime, message handling, transcription and admin-reporting related tests.

Testing

• Ran the automated unit test suite with pytest covering billing, provider runtime, AI requests, message handling and transcription; tests were updated to match the new model id and passed successfully.

---

Codex Task