chore: add language setting to speechv2

[dancer]

background

speech model v2 lacked language setting support, with only provider-specific options available inconsistently across providers.

summary

• add standardized language parameter to Speech Model v2 core interface
• add language support with warnings for unsupported providers

verification

• all tests pass including new language-specific examples
• LMNT uses standardized parameter with fallback to provider options
• OpenAI and Hume show appropriate warnings for unsupported language parameter

tasks

• language parameter added to SpeechModelV2CallOptions
• language parameter added to generateSpeech function
• LMNT provider updated to use core language parameter
• warning implementations for OpenAI and Hume providers
• documentation updated with language examples

future work

• consider removing warnings if providers add native language support

[lgrammel]

pls add a changeset for provider