feat(cli): ferrflow schema subcommand — bundle JSON schema in the binary

[BryanFRD]

Problem

Editor tooling for .ferrflow.json depends on a JSON schema that lives at a stable URL. Today it's published on ferrflow.com/schema/ferrflow.json — handy for VS Code / Zed, useless for offline use or pre-commit hooks that can't hit the internet.

Proposal

Bundle the schema into the ferrflow binary via include_str!, and expose a new subcommand:

ferrflow schema [--pretty] [--output <file>]

• Prints the JSON schema to stdout (or --output).
• --pretty formats it.
• Exits non-zero if the bundled schema can't be parsed (sanity check on the build artefact).

Additionally:

• ferrflow init (feat(init): auto-discover monorepo packages from workspace config #333) uses the bundled schema to validate the generated file right after writing.
• ferrflow doctor (feat(cli): ferrflow doctor diagnostic subcommand #363) uses it too for its "config parses" check.
• Every new FerrFlow release publishes the same schema at a versioned URL, e.g. ferrflow.com/schema/ferrflow-v3.json, so editor integrations can pin to a specific major.

Build-time hook

Add a build.rs (or keep it simple with include_str!) that embeds schema/ferrflow.json from the repo root. CI gate: compare the embedded schema against schema/ferrflow.json — fail if they diverge so we can't ship a binary whose schema doesn't match the checked-in source of truth.

Acceptance

• ferrflow schema returns the schema with exit 0.
• A user with no internet can run ferrflow schema > /tmp/schema.json then point VS Code at that local path.
• Pre-commit hooks in CI can validate .ferrflow.json without any network call.
• The bundled schema matches schema/ferrflow.json byte-for-byte.