Fix first-run setup and startup diagnostics

[pauloacribeiro92]

Summary

• resolve microphone configuration relative to the project root and create its parent directory when needed
• tolerate missing or malformed local microphone configuration
• lazy-load audio and Vosk dependencies so setup helpers can be tested without audio hardware
• report an actionable error when the Vosk model is missing and support BONZI_VOSK_MODEL
• preserve the existing project documentation while correcting the complete first-run procedure
• add standard-library regression tests for configuration and model-path handling

Validation

• added regression tests covering missing configuration, malformed JSON, nested configuration-directory creation, configuration round trips, missing Vosk model diagnostics, and BONZI_VOSK_MODEL overrides
• python -m unittest discover -s tests: passed
• all changed Python files compile successfully
• the previously broken microphone configuration path is now resolved relative to the project root and no longer depends on a nonexistent Windows-only config\mic_config.json location
• missing Vosk model startup now fails with an actionable message instead of surfacing an opaque library error

AI assistance

OpenAI GPT-5.6 Thinking assisted with repository inspection, implementation, and test drafting. Paulo Ribeiro reviewed the scope and remains responsible for the contribution.

Fixes #7
/claim #7

[opirebot]

👀 We've notified the reward creators here.
Make sure your payment account is ready to receive the payment for your hard work 💪