Create a subclass of MCPToolset to simplify connections to Google Managed MCP servers

Base MCPToolset can be tricky to configure when connecting to the newly announced Google Managed MCP servers. The primary complexity is the need to dynamically refresh authentication token obtained using the default application credentials. While it can be done using [header provider](https://github.com/google/adk-python/blob/4f54660d6de54ddde0fec6e09fdd68890ce657ca/contributing/samples/mcp_dynamic_header_agent/agent.py#L31), this feature is not documented is not well known. Also, incapsulating the actual URLs of the MCP servers can improve usability of the toolset. For example,
```
bigquery_toolset = GoogleManagedMCPToolset(Product.BigQuery)
```
would be all that's needed to access the default BigQuery toolset in production environment (deployed on a managed GKE/CloudRun/Agent Engine). Obviously, additional parameters might be needed to configure/filter the tools, but the complexity of authentication and access would be hidden from the developer. It should also simplify integration of with the UI-based Agent Builder (no need to configure complex auth providers).