Add Metadata to Message Class for Displaying Actual User Input

[jrking4]

Currently, when working with chat histories, user requests are sometimes embedded within larger prompt templates. For example:

You are a helpful agent. Please respond to the user request
** User Request **
{{ $UserRequest }}

Please respond only with the answer, no other text is required

In these cases, retrieving the original user message for display or logging purposes requires parsing out the user's text from the overall prompt, which is error-prone and cumbersome. I also see other cases where it maybe useful to get/set metadata to messages.

This PR updates the Message class to support fetching metadata, it is a protected property.

[ilvalerione]

Thank you for the proposal. What about the other use cases? The more you can extend the scenario the better.

[ilvalerione]

I believe this is more an application problem rather than a lack of Neuron.