Cache TUI mark context across batch actions

[jmqd]

Approving/commenting many TUI blocks repeats git discovery, HEAD snapshots, dirty-file checks, and per-path HEAD block reads. Audit refs: trueflow/src/commands/tui.rs:3673, trueflow/src/commands/mark.rs:134, trueflow/src/commands/mark.rs:149, trueflow/src/vcs.rs:244, trueflow/src/vcs.rs:287. Trigger: holding approve across a large review. Direction: let the TUI own a memoized mark context and pass resolved repo metadata/block state into persistence.

[jmqd]

Patch: 0c17d79 reduces mark target-kind inference from up to two scan/tree builds to one scan/tree build by sharing the scanned tree across path+line and hash-only inference. Verification: cargo test target_kind_inference --lib; cargo test commands::mark::tests --lib; just check passed with 994 tests, clippy, and fmt. Remaining issue scope: TUI still needs memoized mark context across repeated actions.

[jmqd]

Patch: c8fc2d3 removes the TUI pre-mark terminal suspend requirement lookup. TUI now calls the noninteractive mark path first and falls back to terminal suspend only on the existing noninteractive signing failure marker, avoiding one extra git config/repo discovery before each TUI mark action. Verification: cargo test execute_mark_for_tui --lib; cargo clippy --features tui-test-support --lib --bins --tests -- -D warnings; just check passed with 994 tests, clippy, and fmt. Remaining issue scope: TUI still needs a reusable mark execution context for repo snapshot, dirty set, and per-path HEAD block caches across repeated actions.