fix(core): reconcile dangling tool_use on interrupt before persisting AgentState (#2410)

Closes #2409.

When an interrupt is signalled between the **reasoning** step (the model
has just emitted `ToolUseBlock`s, which are written into the context at
`runPostReasoningPipeline`) and the **acting** step (tool execution,
which would produce the matching `ToolResultBlock`s), `handleInterrupt`
appended its recovery message without reconciling the pending tool call.