feat: allow plugins to set custom patterns for dangerous commands

[ajgray-stripe]

Description

Refactors the is_dangerous_command utility function to move it to a separate module, then adds a method to that module which allows adding custom patterns. Our use case is that we have some internal CLI tooling touching production resourcing that is definitely dangerous, so we'd like to be able to add an invocation to our plugin to flag those commands as requiring human approval.

Testing

• Added a unit test
• Verified that adding a pattern works by adding a dummy .* entry from the __init__.py of our custom plugin and seeing a command blocked:

[michaelneale]

oh I like this idea