Add a persistent analysis cache to Vulture

Normalized profile

What this task asks for

Add a persistent cache so Vulture can reuse unchanged analysis across runs. Vulture scans every file from scratch on every run, making it slow on large codebases where only a few files changed. `--cache` and `--cache-clear` flags are added to the CLI, with an optional `--cache-dir=PATH` (default `.vulture-cache/`). `--cache-clear` removes all contents of the cache directory before running. The vulture package version must be obtained via `importlib.metadata.version`, and `importlib` must be imported at module scope in `vulture.cache`. Concurrent vulture processes must not corrupt the cache. On every successful save, both a backup of the cache (`cache.json.bak`) and a metadata hash file (`cache.json.meta`) must be written, even on the very first save.