Compile feedback from agentic use-cases into a proposal

[mrm1001]

Write a proposal with suggestions of abstractions that need to be added to help our users create agentic apps. Use the feedback provided on these two use cases:

• https://github.com/deepset-ai/haystack/issues/7738
• https://github.com/deepset-ai/haystack/issues/7740

In particular, some open questions of interest that might come up during the feedback:

• Can we improve how tools are being defined today?
• Do we need to add any abstraction to help our users who want to build CoT/React agents?
• Should we add an OutputValidator component to the library?

This issue can be closed once the proposal has been reviewed and if applicable, follow-up issues have been groomed and added to the board.