feat: add MiniMax-M2.7 as new flagship model

[octo-patch]

Summary

• Add MiniMax-M2.7 as the new flagship model entry (Frontier tier, 1M context window, vision + tools support)
• Update the default minimax alias to resolve to MiniMax-M2.7 (previously pointed to M2.5)
• Add minimax-m2.7 alias for explicit model selection
• Add M2.7 pricing in the metering module
• Update model catalog tests for M2.7 as the new default
• Increment MiniMax model count from 6 to 7

All existing M2.5, M2.5-highspeed, M2.1, and legacy abab models remain available via their existing aliases.

Test plan

• cargo check -p openfang-runtime -p openfang-kernel passes
• All 43 model catalog unit tests pass (cargo test -p openfang-runtime -- model_catalog)
• Metering tests pass
• Verify minimax alias resolves to MiniMax-M2.7
• Verify minimax-m2.7 alias works
• Verify existing minimax-m2.5 alias still works

[jaberjaber23]

4 of 5 model specifications are wrong:

1. Pricing: Listed $1.10/$4.40 per 1M tokens. Actual MiniMax M2.7 pricing is $0.30/$1.20 (3.7x inflated).
2. Context window: Listed 1,048,576 (1M tokens). Actual is 200K-204.8K tokens.
3. Max output: Listed 16,384. Actual is 131,072.
4. Vision: Listed supports_vision: true. M2.7 is text-only, no vision support.

All values must be corrected against the official MiniMax API documentation before this can be merged. The metering.rs rates also need to match.