refactor(syntax): consume published Syntaxmate crate

[phongndo]

Summary

• replace Mark's in-tree Rust-native TextMate implementation with the published syntaxmate 0.1.0 registry crate
• retain crates/mark-syntax as a thin product adapter for settings, language enablement, exact-scope rendering, and Mark's 63-theme catalog
• remove duplicated grammar assets, engine/oracle code, fixtures, generated reports, and obsolete CI shards

Motivation

• Problem: the engine, catalog, and compatibility harness were coupled to Mark's repository and release cycle.
• Why now: Syntaxmate 0.1.0 is independently published, release-attested, and validated; Mark can now exercise the same public API as any downstream consumer.

Changes

• resolve syntaxmate = "0.1" exclusively from crates.io with bundled-grammars; benchmark-only diagnostics remain feature-gated
• adapt Syntaxmate tokenizer output into Mark's existing line/segment model while preserving exact shared scope tables and coarse fallback classes
• delegate TextMate selector parsing/resolution to Syntaxmate while keeping Mark's TUI theme assets and rendering policy
• move tokenizer oracle/fuzz/generated-grammar ownership to Syntaxmate and retain downstream rendering, queue, settings, packaging, and performance gates in Mark
• add architecture checks that reject copied engine/assets, internal Syntaxmate modules, and path/Git/patch resolution
• package the required Syntaxmate grammar notices with Mark distributions

Verification

• scripts/ci/rust
• scripts/ci/generated
• scripts/ci/performance extended
• cargo +1.88.0 check --workspace --all-targets --all-features --locked
• cargo test --workspace --all-targets --all-features --locked (via scripts/ci/rust)
• scripts/check-architecture
• actionlint -color
• isolated distribution archive/checksum and all Syntaxmate notice names
• cd pi-mark && pnpm run check (unchanged; selected by CI because workflow orchestration changed)

Risk

• Blast radius: syntax tokenization, exact-scope theme resolution, language detection, packaged notices, and syntax benchmark instrumentation.
• Rollback: revert this commit to restore the prior in-tree engine and assets; no user configuration migration is required.

Reviewer Notes

• Review focus: registry-only dependency/lockfile source, adapter scope-table ownership, diagnostics feature isolation, 63-theme preservation, and removal of duplicate engine assets/tooling.

[Copilot]

Copilot wasn't able to review this pull request because it exceeds the maximum number of files (300). Try reducing the number of changed files and requesting a review from Copilot again.