Feat/flux support

[dpmishler]

Add Flux Conversational Speech Recognition Support

This PR adds support for Deepgram's Flux model (flux-general-en), the first conversational speech recognition model built specifically for voice agents.

Features Added

• New flux_request() and flux_request_with_options() methods for Flux streaming
• Support for turn-based conversation detection with FluxResponse types
• Configurable end-of-turn detection parameters:
  • eot_threshold - Confidence required for EndOfTurn events
  • eager_eot_threshold - Confidence for early EagerEndOfTurn events
  • eot_timeout_ms - Maximum silence before forcing turn end
• New TurnEvent enum with variants: StartOfTurn, EndOfTurn, EagerEndOfTurn, TurnResumed, Update
• Uses /v2/listen endpoint for Flux API

Examples

• simple_flux - File streaming example
• microphone_flux - Real-time microphone streaming example

Implementation Details

• Follows existing websocket.rs patterns for consistency
• Comprehensive error handling and edge case coverage
• Partial frame handling for fragmented JSON messages
• Proper connection state tracking and cleanup
• Tests for URL construction and query parameter encoding

Documentation

• Updated CHANGELOG.md with Flux feature details
• Updated examples/README.md with Flux examples
• Inline documentation with links to Deepgram Flux API Reference

Testing

• All existing tests pass (131 tests)
• New tests for Flux URL construction and query encoding
• Verified with real API using both file and microphone examples