fix(faster-whisper): cast segment timestamps to int after multiplication

[arteven]

Summary

Every audio transcription request through the faster-whisper backend currently fails with:

TypeError: 'float' object cannot be interpreted as an integer

Root cause

In backend/python/faster-whisper/backend.py:

backend_pb2.TranscriptSegment(id=id, start=int(segment.start)*1e9, end=int(segment.end)*1e9, ...)

int(x) * 1e9 produces a float because 1e9 is a float literal in Python. The protobuf TranscriptSegment.start / .end fields are integers, so the gRPC call raises a TypeError and the whole request fails.

Fix

Multiply first, then cast to int — int(x * 1e9). One-character-positions changed.

Test plan

• Reproduced the bug locally with Systran/faster-whisper-tiny on v4.1.3 — every request returns 'float' object cannot be interpreted as an integer.
• Verified the patched backend transcribes audio successfully end-to-end via /v1/audio/transcriptions.
• CI passes.