feat(minimax): add speech audio generation

[octo-patch]

Reason: Enable MiniMax speech-2.8 text-to-speech through Rig's audio-generation client abstraction.

Adds feature-gated audio generation to the OpenAI-compatible MiniMax client for speech-2.8-hd and speech-2.8-turbo.

The implementation sends requests to /v1/t2a_v2, maps the shared text and voice request fields to MiniMax's speech schema, decodes hexadecimal data.audio, and preserves HTTP and provider-envelope errors.

Checks:

• cargo fmt --all -- --check
• cargo test -p rig-core --features audio providers::minimax::audio_generation --lib
• cargo test -p rig-core --features audio minimax --lib
• cargo test -p rig-core --features audio --lib
• cargo clippy -p rig-core --features audio --lib --tests -- -D warnings
• cargo test -p rig --all-features --test minimax