Fix MCP test independence

[daniel-lxs]

Ensures each MCP test sets up its own configuration independently by adding alwaysAllow permissions setup to all tests. This prevents test failures when tests run in different orders.

---

Important

Ensures MCP test independence by setting up individual configurations and introduces mcpServersInitialized event for server initialization handling.

• Behavior:
  • Ensures each MCP test in use-mcp-tool.test.ts sets up its own configuration independently by modifying the mcpConfig to include alwaysAllow permissions.
  • Adds logic to modify and save the MCP config file to trigger server detection in use-mcp-tool.test.ts.
  • Introduces a 1-second delay after modifying the config to allow server initialization.
• Events:
  • Adds mcpServersInitialized event to RooCodeAPIEvents in api.ts and ipc.ts.
  • Emits mcpServersInitialized in McpHub after server connections are updated.
• Misc:
  • Extends ClineProviderEvents in ClineProvider.ts to include mcpServersInitialized.
  • Updates API in api.ts to listen for mcpServersInitialized and emit it.

^{This description was created by for e52478f. You can customize this summary. It will automatically update as commits are pushed.}

[daniel-lxs]

This seems to help @cte @mrubens

[cte]

Could we hook into our event emitter to determine if we're initialized instead of sleeping?

[daniel-lxs]

It seems like taskStarted will not wait for MCP servers to be initialized causing the model to not see them.

Do you know of any event I can use to wait for the MCP servers to initialize?

interface RooCodeAPIEvents {
    message: [data: {
        taskId: string;
        action: "created" | "updated";
        message: ClineMessage;
    }];
    taskCreated: [taskId: string];
    taskStarted: [taskId: string];
    taskModeSwitched: [taskId: string, mode: string];
    taskPaused: [taskId: string];
    taskUnpaused: [taskId: string];
    taskAskResponded: [taskId: string];
    taskAborted: [taskId: string];
    taskSpawned: [parentTaskId: string, childTaskId: string];
    taskCompleted: [taskId: string, tokenUsage: TokenUsage, toolUsage: ToolUsage, isSubtask: IsSubtask];
    taskTokenUsageUpdated: [taskId: string, tokenUsage: TokenUsage];
    taskToolFailed: [taskId: string, toolName: ToolName, error: string];
}

[cte]

Maybe we should add one?

[delve-auditor]

We have finished reviewing your PR. We have found no vulnerabilities.

Reply to this PR with @delve-auditor followed by a description of what change you want and we'll auto-submit a change to this PR to implement it.