Add MiniMax M3 to provider catalog

[octo-patch]

Summary

• Register MiniMax-M3 in the built-in catalog and make it the default MiniMax model across setup flows.
• Keep MiniMax-M2.7 available while correcting its output limit from the current official specification.
• Preserve M3 input-length pricing and document regional OpenAI-compatible and Anthropic-compatible configuration.

Changes

• Add M3 context, output, vision, tool, alias, and pricing metadata.
• Apply the long-context standard rate when input exceeds 512,000 tokens.
• Add a request-capture test for the Anthropic-compatible /v1/messages path.
• Document both regions, both protocols, thinking modes, modalities, and all current pricing tiers.

Testing

• cargo fmt --all -- --check
• cargo test --workspace
• cargo clippy --workspace --all-targets -- -D warnings
• Live integration tested (not run; local request capture verifies endpoint construction without external credentials)

Security

• No new unsafe code
• No secrets or API keys in diff
• User input validated at boundaries