feat(test): Evaluate MCP tools usage

[MichalKalita]

Closes https://github.com/apify/ai-team/issues/30

It's a simple test to run Agent with a prompt and check how the tools were used.
The target is to be sure the tool description is good and LLM doesn't make mistakes with tool selection and arguments.

[MichalKalita]

This MCP server is not compatible with the Mastra MCP client. We allow adding tools (Actors) at runtime inside LLM requests, while Mastra only allows adding tools between requests. This may be a problem for more clients.

We have two options:

1. Implement Add generic call-actor tool #155
2. Use on MCP client other than Mastra ( but it's itself a problem, when popular MCP clients don't work, it's itself an evaluation fail)

@jirispilka @MQ37

[MQ37]

This MCP server is not compatible with the Mastra MCP client. We allow adding tools (Actors) at runtime inside LLM requests, while Mastra only allows adding tools between requests. This may be a problem for more clients.

We have two options:

1. Implement [Add generic call-actor tool #155](https://github.com/apify/actors-mcp-server/issues/155)

2. Use on MCP client other than Mastra ( but it's itself a problem, when popular MCP clients don't work, it's itself an evaluation fail)

@jirispilka @MQ37

I think this conclusion makes sense, I think we should implement the generic call-actor tool 👍

[jirispilka]

@MichalKalita is this PR still relevant since we have generic call-actor tool

[MichalKalita]

@jirispilka I'm closing this PR. We want a tool that allows us to conduct A/B tests in complex scenarios, collect all metrics, and decide which way is better.