fix: fall back when ripgrep refresh fails

[yosofbadr]

Summary

Fixes #1338 by making directory refresh call the detected rg executable directly and fall back to the internal scanner if ripgrep exits unsuccessfully.

The previous shell-string invocation could fail to use the detected ripgrep path correctly. This switches to an argument-list subprocess call, cleans up the generated ignore file in a finally block, and adds regression coverage for the ripgrep-failure fallback path.

Tasks Completed

• Platforms Tested:
  • Windows x86 - GitHub Actions pytest passed; not manually desktop-tested
  • Windows ARM
  • macOS x86
  • macOS ARM - targeted local test/lint
  • Linux x86 - GitHub Actions pytest passed
  • Linux ARM
• Tested For:
  • Basic functionality - targeted refresh fallback regression test
  • PyInstaller executable
    • Local: uv run --extra pytest python -m pytest tests/macros/test_refresh_dir.py -q
    • Local: uv run --extra ruff ruff check src/tagstudio/core/library/refresh.py tests/macros/test_refresh_dir.py
    • CI: Ruff, Pyright, REUSE, pytest Linux, pytest Windows, coverage all passed.
    • I did not manually reproduce the original Windows desktop setup with ripgrep installed.

[yosofbadr]

Closing this after a stricter look. This issue is specifically about refreshing on Windows with ripgrep installed, and I have not tested that actual Windows setup. The added regression test only covers the fallback path, so I do not want to leave this open as if it were fully validated.

Sorry for the noise.