[Python] Audit log data model and storage (read/write _fd5_audit_log)

[gerchowl]

Parent: #161

Goal

Create fd5.audit module with AuditEntry dataclass and read/write functions for the _fd5_audit_log root attribute.

Spec

• AuditEntry dataclass: parent_hash, timestamp (ISO-8601), author (dict with type/id/name), message, changes (list of dicts with action/path/attr/old/new)
• read_audit_log(file: h5py.File) -> list[AuditEntry] — parse JSON from _fd5_audit_log attribute
• append_audit_entry(file: h5py.File, entry: AuditEntry) — read existing log, append, write back as JSON
• AuditEntry.to_dict() / AuditEntry.from_dict() for JSON serialization
• The _fd5_audit_log attribute is a VarLenUnicode string (JSON array)
• NOT excluded from content_hash computation (tamper-evident)

TDD

• Test round-trip: write entry → read back → assert equal
• Test append to existing log
• Test empty log returns []
• Test malformed JSON raises clear error
• Test entry validation (required fields)