Add possibility to check if a tool was run

Problem/Motivation

While #3548000: Add possibility to run a tool at a specific order → would take care of forcing providers to use a tool, sometimes we might not know when we want a tool to run, but we do know that the agent should run it before its finished.

An example would be for instance a validation agent, even if the instructions are completely off what it should be doing, should still use the validation tool to decline the instructions.

We can do this outside of provider methods, by having a deterministic check at the end of the agent run and if the tool was not used, we do not end the agent, but add another user message with a predfined configurable message.

Proposed resolution

• Add a configuration per tool that you can force it to have run.
• Add a configuration of the instruction to send if it didn't run, but the agent still assumes it is finished.
• On the runner have a check that all configured "force to run" tools have run, if not we append a user message and start another loop.