refactor: harden codebase and remove unicode artifacts

- fix urllib.parse.quote import (was urllib.request.quote)
- add operator precedence parens in generate_pack dedup check
- narrow bare except to specific types in batocera target scraper
- cache load_platform_config and build_zip_contents_index results
- add selective algorithm support to compute_hashes
- atomic write for fetch_large_file (tmp + rename)
- add response size limit to base scraper fetch
- extract build_target_cores_cache to common.py (dedup verify/pack)
- hoist _build_supplemental_index out of per-platform loop
- migrate function-attribute caches to module-level dicts
- add @abstractmethod to BaseTargetScraper.fetch_targets
- remove backward-compat re-exports from common.py
- replace em-dashes and unicode arrows with ASCII equivalents
- remove decorative section dividers and obvious comments

scripts/scraper/targets/batocera_targets_scraper.py

@@ -136,7 +136,7 @@ def _check_atom(tokens: list[str], pos: int, active: frozenset[str]) -> tuple[bo
     if tok.startswith('"'):
         pos += 1
         return True, pos
-    # Unknown token — treat as true to avoid false negatives
+    # Unknown token -treat as true to avoid false negatives
     pos += 1
     return True, pos