feat: separate core workflows and add TUI-lite terminal UX

[Teakowa]

Summary

Implements the command-boundary and human terminal UX work tracked by #209 and #210.

• check is now a correctness/frontend/semantic/validation gate and no longer emits ordinary configurable lint findings.
• lint remains the configurable rule workflow with stable rule metadata, evidence, boundedness, severity, and findings.
• analyze now reports semantic structure, symbol usage, and CFG measurements through result.facts, rather than exposing the lint registry.
• Interactive terminal text gets delayed TTY-only activity feedback, command-specific PASS/WARN/ERROR summaries before details, progressive styling, and source context where the provenance path is readable.
• JSON, plain, redirected, CI, GitHub Actions, and TERM=dumb paths remain static and machine-safe.

Contract changes

The wright-result/v1 envelope, exit codes, and JSON purity remain stable. The AnalyzeResult payload changes from result.findings to result.facts; the new facts contain deterministic symbol usage and per-rule CFG measurements. Lint findings remain under result.findings for lint and the existing agent/tool query surfaces.

Validation

• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• cargo test --workspace --all-targets --all-features
• Fresh representative CLI checks for JSON, plain, TERM=dumb, and pseudo-TTY execution.

This branch is stacked on PR #208 and should be reviewed with ci/reorganize-by-capability-207 as its base.

Fixes #209
Fixes #210