[Feature Request] - Support for task chaining and multi-agent conversations with shared memory

[dannleft]

Description

Is your feature request related to a problem? Please describe.
Currently, there is no clear or documented way to:

1. Chain outputs and inputs when tasks depend on the results of other tasks, including parallel task flows.
2. Set up multi-agent interactions where agents share a memory of their interactions to provide context for subsequent tasks.

This makes it challenging to create complex workflows or conversational agents that rely on shared context or parallel task resolution.

Describe the solution you'd like
I would like KaibanJS to provide:

1. A formalized mechanism for defining task dependencies, including chaining outputs as inputs and handling tasks running in parallel.
2. Support for multi-agent interactions where:
   • Agents can exchange information in a 'ping-pong' conversational flow.
   • A shared memory or state manager allows agents to retain and access the entire sequence of conversations as context for further processing.
3. Clear examples and documentation for these features.

Describe alternatives you've considered

• Manually managing task dependencies and outputs outside KaibanJS, which introduces complexity and reduces the efficiency of the framework.
• Using an external state management tool to handle shared memory for agents, which requires significant integration effort.

Additional context
This feature would enable developers to create advanced use cases like:

• Complex task workflows where outputs dynamically feed into other tasks.
• Multi-agent simulations where agents collaboratively solve problems or reach agreements.
• Real-world AI applications like collaborative bots or automated negotiation systems.