feat(tesla): raise PrivateKeyError for unusable existing key files - #146

Intent

Review-comment fix on PR 146: PrivateKeyError must subclass LibraryError, not TeslaFleetError, since a key-file failure is local to this library, not an upstream Fleet API error. Moved the class next to SigningDisabled under LibraryError in exceptions.py, kept the reason/message attributes and docstring, updated the docstring's rationale and the README line that described it as a TeslaFleetError subclass. LibraryError subclasses Exception (not BaseException); PrivateKeyError(...) from err raise sites in tesla.py are unaffected by this base-class change. Tests still assert PrivateKeyError itself and needed no changes.

What Changed

• Added PrivateKeyError (a LibraryError subclass, not TeslaFleetError, since a key-file failure is local to this library rather than an upstream Fleet API error) with a reason of "unreadable", "malformed", "encrypted", or "wrong_type", plus a message attribute.
• Tesla.get_private_key/get_rsa_private_key now route a known-existing key file's read/parse failure through a shared _load_existing_private_key helper that raises PrivateKeyError for I/O errors, invalid PEM, password-encrypted PEM, and a loaded key of the wrong type, replacing the previous ad-hoc FileNotFoundError/PermissionError/AssertionError handling; key generation and the O_EXCL create-race fallback are unaffected.
• Updated the README and docs/bluetooth_vehicles.md/docs/fleet_api_energy_sites.md to document the new PrivateKeyError reasons and its LibraryError base, consolidating the create/load semantics description in the README with cross-references from the docs.
• Added PrivateKeyErrorTests covering unreadable, malformed, encrypted, and wrong-type key files for both the EC and RSA loaders.

Risk Assessment

✅ Low: Small, well-scoped change: moves PrivateKeyError from TeslaFleetError to LibraryError, correctly threads the message through Exception.init (required since LibraryError has no custom init unlike TeslaFleetError), updates docs consistently, and has no internal call sites that catch TeslaFleetError/BaseException expecting to intercept PrivateKeyError, so no behavior regression.

Testing

Ran the existing targeted unit tests in tests/test_tesla_private_key.py (all 38 pass unchanged, confirming the intent's claim that no test changes were needed) and additionally wrote a manual end-to-end script exercising the actual exception-hierarchy behavior change: it proves a caller's generic except TeslaFleetError handler no longer swallows PrivateKeyError post-fix, while except LibraryError does, and that reason/message/str() remain intact. No issues found; working tree left clean.

OK: propagated past except TeslaFleetError -> reason='malformed' message='Private key file at /tmp/.../key.pem is not a valid PEM private key'
OK: caught by except LibraryError -> reason='malformed'
OK: str(e)='Private key file at /tmp/.../key.pem is not a valid PEM private key'

Pipeline

Updates from git push no-mistakes