…minals

Adds /subtask [--last N] <prompt> command that spawns a new codex instance
in a separate terminal window for parallel exploration and task execution.

- Implements terminal detection and spawning for Linux (Ghostty, VS Code,
  Alacritty, Kitty, WezTerm, gnome-terminal, konsole, xterm)
- Non-blocking operation allows main session to continue independently
- Passes current working directory to subtask
- Adds foundation for passing conversation history (--last N messages)
- New spawn_terminal module handles cross-terminal compatibility

The command is available during task execution and useful for exploring
alternative approaches or working on related tasks in parallel.

Allows users to specify which model to use when spawning a subtask in
a new terminal. The model flag can be combined with --last in any order.

Examples:
- /subtask --model claude-opus-4 Fix authentication
- /subtask --last 5 --model gpt-5-codex Refactor module

Changes:
- Enhanced parse_subtask_args() to parse --model flag
- Updated InputResult::SubtaskCommand with model field
- Updated AppEvent::SpawnSubtask with model field
- Modified handle_spawn_subtask() to pass -m flag to spawned codex
- Updated help text with --model documentation

- Changed binary name from 'codex' to 'vcodex' in cli/Cargo.toml
- Changed default config directory from ~/.codex to ~/.vcodex
- Updated documentation to reflect new binary and config names
- This allows vcodex fork to coexist with vanilla codex installation

…session

- Added /rejoin slash command that reads the most recent session
- Extracts and displays last 2-3 assistant messages from subtask
- Formats results as "Subtask <id> results: <summary>"
- Made get_conversations public in codex-core for TUI access
- Integrates with existing rollout reading infrastructure