refactor: port planner identities, scopes, rules and CLI runtime to Rust

[kvlonge]

Why

Continue the Rust-hosted compiler, CLI and planner migration while preserving compiled artifacts, persisted identities and command behavior.

Changes

• Native planner identities, pruning, iterative deep-graph traversal, selectors, scopes and loader graphs. Existing Python executor and adapter boundaries remain.
• Native rule catalogue, guidance, run planning, SQL-rule execution and early lint; retain custom-rule execution and its public fact views in Python.
• Native CLI setup, observability, adapter resolution, warehouse scope, plan/rules commands, compile reuse, replay and launch-record writing.
• Native janitor phases, runtime artifacts and coded-error rendering.
• Stop and join live progress spinners before interpreter finalization.
• Restore linear SQL-rule cache publication by grouping findings once; preserve ordering and cache format.
• Movable Python decreases from 51,236 to 42,632 lines, including agreed user-code/public-API boundary classifications.

Verification

• 11,391 unit/integration tests passed; 86 focused tests after the cache-publication fix; updated make check-ci passed.
• Compiler differential: 371 projects with discovery/render coverage. Stateful planner differential: all 11 projects matched the existing baseline across 18 steps.
• Cold and edited compile outputs and exit codes match the reference on synthetic and private compile-only projects.
• Generated identity, selection and scope comparisons; pinned persisted hashes; 100,000-node graph regression tests.
• Shutdown probe reproduced 5/5 aborts before the fix and 20/20 clean exits afterward; disabling the hook reproduced the abort.
• Independent review found quadratic cache publication; fixed with fail-before/pass-after operation-count coverage and verified in a bounded follow-up review.