Fix CPU sidecar builds bundling CUDA torch instead of CPU

The CPU build steps used `uv run pyinstaller` which re-resolves
dependencies from pyproject.toml's [tool.uv.sources] before running,
pulling CUDA torch back in after the CPU-only reinstall. This made
CPU and CUDA zips the same size.

Fix: run pyinstaller directly from the venv (.venv/bin/pyinstaller
on Linux/macOS, .venv\Scripts\pyinstaller.exe on Windows) to skip
uv's dependency resolution entirely.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.gitea/workflows/build-sidecar.yml

@@ -152,7 +152,9 @@ jobs:
         run: |
           rm -rf dist/local-transcription-backend build/
           uv pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu --force-reinstall
-          uv run pyinstaller local-transcription-headless.spec
+          # Run pyinstaller directly from venv to prevent uv run from
+          # re-resolving torch back to the CUDA version via pyproject.toml sources
+          .venv/bin/pyinstaller local-transcription-headless.spec
 
       - name: Package sidecar (CPU)
         run: |
@@ -268,7 +270,9 @@ jobs:
         run: |
           Remove-Item -Recurse -Force dist\local-transcription-backend, build -ErrorAction SilentlyContinue
           uv pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu --force-reinstall
-          uv run pyinstaller local-transcription-headless.spec
+          # Run pyinstaller directly from venv to prevent uv run from
+          # re-resolving torch back to the CUDA version via pyproject.toml sources
+          .venv\Scripts\pyinstaller.exe local-transcription-headless.spec
 
       - name: Package sidecar (CPU)
         shell: powershell
@@ -370,10 +374,9 @@ jobs:
         run: |
           # UV_NO_SOURCES bypasses pyproject.toml's [tool.uv.sources] which forces
           # torch from the CUDA index (no macOS ARM wheels there).
-          # Applies to both uv sync AND uv run (which re-resolves).
           # Default PyPI torch includes MPS (Apple Silicon GPU) support.
           uv sync
-          uv run pyinstaller local-transcription-headless.spec
+          .venv/bin/pyinstaller local-transcription-headless.spec
 
       - name: Package sidecar (CPU)
         run: |