Add plugin: `preflight-disk-space` (abort below free-space threshold)

Agents that write reports, dumps, or logs can fail mid-run with ENOSPC, leaving partial state. The examples/disk-alert/ agent already monitors disk space, but there is no preflight that prevents an agent from starting when disk is below a threshold.

Proposal

A new plugin preflight-disk-space that aborts an agent run if free space on a target volume is below a configurable threshold.

Manifest usage:

[[preflight]]
plugin = "preflight-disk-space"
path = "/Users/me/reports"
min_free_gb = 5            # abort if free space < 5 GiB
# or:
min_free_percent = 10      # abort if less than 10% free

Acceptance criteria

• New plugin crate under plugins/preflight-disk-space/ following the convention of preflight-cmd.
• Both absolute (min_free_gb) and percentage (min_free_percent) thresholds work; either or both.
• When both are set, abort if either is breached.
• suggest field on abort names which volume is full and how much short it is.
• Works on macOS and Linux (use statvfs or equivalent cross-platform crate).
• Doc page under docs/plugins/preflight-disk-space.md.
• Registered in the workspace Cargo.toml.

Where to start

• plugins/preflight-cmd/ — convention to copy.
• docs/reference/plugin-protocol.md — protocol.
• examples/disk-alert/ — the existing reactive disk agent for inspiration.

Non-goals

• Inode exhaustion checks (separate failure mode, future plugin).
• Mount-point health beyond free space.