fix: missing f-string prefix and temp file cleanup in voice.py

[gambletan]

Summary

• Fix missing f prefix on warning string (line 149) — currently prints literal {temp_wav} instead of the variable value
• Replace deprecated tempfile.mktemp() with tempfile.mkstemp() to fix TOCTOU race condition
• Add proper temp file cleanup to prevent file leaks on both success and error paths

Test plan

• Verify f-string now interpolates correctly
• Verify temp files are cleaned up after transcription
• Verify behavior unchanged for both wav and mp3 paths

🤖 Generated with Claude Code

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}