runQuery hook threaded through tool-execution contexts

What

• The isolated tool-execution context used to expose Claude Code's tools over an MCP server connection now includes a runQuery function in its options.
• The function that builds options for a non-interactive/analysis query now accepts a runQuery parameter from its caller and passes it into the agent context options, alongside existing options like commands, tools, and mcpClients.
• The interactive REPL's tool-use context now also includes a runQuery entry, alongside existing options like resolveCommandDialog and renderDebugControls.

Why Giving all three tool-execution contexts (MCP bridge, non-interactive queries, and the interactive REPL) the same runQuery hook lets them share a consistent way to run queries, rather than each context wiring this up separately or lacking the capability.