Add support for Claude Code's `--dangerously-execute-all` flag

[Escoto]

Problem

Claude Code has a powerful flag called --dangerously-execute-all that allows it to automatically run shell commands (package installation, file editing, running tests, git operations, etc.) without asking for confirmation every time.

Currently, RustifyMyClaw does not expose this flag, forcing users to manually approve every command when using it through Telegram, WhatsApp or Slack. This significantly reduces the agentic experience many users want.

Proposed solution

Add support for the --dangerously-execute-all flag at the config YAML level, so it can be enabled per workspace.

# ~/home/user/.rustifymyclaw/config.yaml
workspaces:
  default:
    backend:
      name: claude-code
      dangerously_execute_all: true     # default: false
      # other options...

  # Can be enabled only for specific workspaces
  personal:
    backend:
      name: claude-code
      dangerously_execute_all: false

  work:
    backend:
      name: claude-code
      dangerously_execute_all: true

Additional context

• This should follow the same pattern used by other config options in config.rs and the backend implementations.
• Update the configuration examples in the README.md (both main example and Claude Code section).
• Add a warning message after the program starts.
• Include a test case for the new option.