feat: enhance load_mcp_tools to support loading specific tool names

[msamon-nrg]

Summary

This PR updates the load_mcp_tools function in tools.py to support loading a specific subset of tools by their IDs. It also adds new unit tests to verify this behavior.

Details

Feature:

The load_mcp_tools function now accepts an optional tool_names argument (a list of string tool names/IDs).
If tool_names is provided and not empty, only tools with matching names are loaded.
If tool_names is empty or not provided, all available tools are loaded (existing behavior).

Tests:

Added test_load_mcp_tools_with_no_tool_names_e2e to verify that all tools are loaded when no IDs are specified.
Added test_load_mcp_tools_with_specific_tool_names_e2e to verify that only the specified tools are loaded when IDs are provided.

Motivation

This change allows users to selectively load only the tools they need, improving efficiency and flexibility when working with large sets of MCP tools.

Additional Notes

All new and existing tests pass with the updated implementation.

[eyurtsev]

Overall looks good. I'd like to move away from unit tests into end to end tests since the abstraction is so thin.

[msamon-nrg]

Ok @eyurtsev are there any examples of end to end tests in the repo already?

[eyurtsev]

Yes -- git grep for with run_streamable_http(server) -- it's in the same file as you've added tests.

We'll also need to replace tool_ids with tool_names to match the mcp spec.

[msamon-nrg]

Conflict resolved. This PR is ready for another review @eyurtsev

[grantmca]

I like this, it would be really nice if we could get a public method for listing MCP tools. That way we could filter on _meta and only convert the tools we want (Or on name)