Define and implement the Ceph-backed volume lifecycle

[halfcyan]

Closes #102

Summary

This stacked PR completes the Ceph-backed storage contract on top of #143, which supplies the shared provider lifecycle machinery. Together, the stack:

• Documents local-file versus Ceph RBD behavior, the supported URI contract, ownership/attachment/mapping/teardown/deletion states, and operator configuration.
• Resolves existing RBD images directly to host block devices; it does not copy images to compute-node-local storage or implicitly create/delete Ceph resources.
• Tracks owned, borrowed, and shared mappings; prevents premature unmapping and makes cleanup retryable/idempotent.
• Uses a mockable storage-provider interface and fake-provider tests, so CI does not require Ceph.
• Validates RBD URI syntax before URL normalization or Cloud Hypervisor handoff.

Pool/image references are supplied by rbd://<pool>/<image> in each manifest attachment; CEPH_CONFIG, CEPH_ID, CEPH_KEYFILE, and CEPH_CLUSTER configure the node-local Ceph client.

Validation

• cargo fmt --all -- --check
• cargo test -p odorobo
• cargo clippy -p odorobo --all-targets -- -D warnings

This PR is stacked on #143; #143 must land first.