Share sessions across machines, and developers

### Describe the feature or problem you'd like to solve

I might work on different machines to solve the same problem. I also have teams working together to solve the same problem. It would be good if discoveries, context, history and failures were shareable across sessions and people.

### Proposed solution

Develop some sort of knowledge pool concept. A "problem space" could be created. This could contain references to multiple git repos, documentation, copilot instructions, subagents, MCP server configs etc etc. If the problem space or knowledge pool was built up over time, it would reduce the effect or feeling that every conversation starts from scratch. "/session" is a partial solution to this on a single developers machine, but it does not seem to scale very well. In terms of cost and resource utilisation, and maybe even model training, building up knowledge and known to be true facts could save lots of compute time and improve UX and satisfaction of development teams. This sort of thing would take copilot from single dev's sidekick or swiss army knife, into a Enterprise productivity solution.

### Example prompts or workflows

Workflow: Developer A defines a subagent, everyone in the team can leverage it.
Workflow: A dev team can work collaboratively on a whole platform, over many git repos. Shared context or even a small model could be generated and maintained that improves and evolves as the application being worked upon evolves.
Workflow: The subagents or personas we create could be orchestrated as a team/group to solve larger problems, covering specific angles, such as QA, SRE, DevOps, Security, etc etc.

### Additional context

_No response_

No implementation files or tests are named. Start by examining the existing /session behavior and how it stores session context locally, then clarify the scope for shared problem spaces across machines and developers. Done requires an agreed design and acceptance criteria for sharing context, repositories, instructions, subagents, and MCP configurations.

Written by the indexing model from the issue text.