Deepgram Flux TTS in EVA

[parthamukherjeesnow]

Changes made to wire Deepgram Flux TTS into eva:

• pyproject.toml / uv.lock: bumped pipecat-ai floor from >=1.4.0 to >=1.6.0, resolved/synced to 1.7.0.
• src/eva/assistant/pipeline/services.py: create_tts_service now dispatches "deepgram-flux" to
  DeepgramFluxTTSService (voice defaults to flux-alexis-en), alongside the existing plain "deepgram" (Aura)
  path — mirroring how deepgram-flux STT is already handled.
• .env.example: added deepgram/deepgram-flux to the TTS enum list plus a config example.
• tests/unit/assistant/test_services.py: added 3 tests for the new dispatch (Flux vs. Aura selection, default
  voice).

Full unit suite (2035 tests) and ruff pass. To use it, set EVA_MODEL__TTS=deepgram-flux with
EVA_MODEL__TTS_PARAMS='{"api_key": "...", "voice": "flux-haley-en"}'.