Show expression to be called in nrepl_eval when asking for permission

Similar to https://github.com/dirge-code/dirge/issues/587 if the `plugin_tool` calls `nrepl_eval`, I can't know what it intends to call because the alert box does not show the expression. I can just disallow it or trust the agent blindly.

Would it be possible to show what expression will be evaluated?

Start by tracing the permission alert shown when the plugin_tool calls nrepl_eval, then follow how the requested evaluation is passed into that prompt. The work is done when the alert displays the expression that will be evaluated, allowing the user to approve or reject it knowingly.

Written by the indexing model from the issue text.