Implement manifest command trust storage and basectl trust commands

[codeforester]

Follow-up to #1367

docs/manifest-command-trust.md defines the model for local manifest command approval. This issue implements the storage and command surface before enforcement is wired into project command execution.

Scope

• Compute manifest command trust identity: project root, manifest path, manifest SHA-256, project name, Git root, sanitized origin, and HEAD metadata when available.
• Store allow records under ~/.base.d/trust/manifest-commands/ with schema version 1 and atomic writes.
• Add basectl trust status <project> [--workspace <path>] [--format text|json].
• Add basectl trust allow <project> [--workspace <path>] [--manifest-sha256 <sha256>].
• Add basectl trust revoke <project> [--workspace <path>].
• Include non-interactive behavior for --manifest-sha256 mismatches.

Acceptance Criteria

• Trust identity and storage are covered by Python unit tests.
• trust status --format json emits the structured payload described in the design doc.
• Allow records are written only under Base-managed local state, never inside project repositories.
• Command help and command reference are updated.