Add generic bb document import MVP

[codeforester]

Summary

Add the first document-first v2 workflow: importing an explicit file through bb documents import without bank-specific parsing or inference.

Scope

• Add a bb documents import --file FILE command.
• Add --dry-run planning that reports hash, size, media type, and canonical object key without writing database rows or copying files.
• On real import, hash the file, create or reuse a BB_DOCUMENT row, copy the source file into financial/canonical, and create a BB_DOCUMENT_OBJECT row.
• Keep canonical objects addressed by SHA-256 object keys, not source filenames.
• Make duplicate imports idempotent by file hash/object key.
• Keep parser detection, inference, views, and document inbox processing out of scope.

Acceptance criteria

• uv run bb documents import --dry-run --file sample.pdf reports a plan and makes no changes.
• uv run bb documents import --file sample.pdf records the document and copies a canonical object.
• Re-importing the same file is safe and reports the existing document/object rather than creating duplicates.
• bb status and existing legacy commands continue to work.
• Validation passes.

Follows #130.