feat: implement multi-agent support and 'use' command

[UmmeHabiba1312]

This PR introduces the ability to manage multiple AI integrations within a single project.

• Updated save_init_options to maintain a list of installed_integrations.
• Modified init to allow adding new agents to an existing project without conflict.
• Added list-integrations command to view all installed agents.
• Added use command to switch between active integrations.

Fixes #2257

[mnriem]

Explain why this is needed? This will muddy the waters for any coding agent as now it has potentially N number of agents / commands and skills in the repository that it sees?

[UmmeHabiba1312]

Thanks for the feedback @mnriem . This is primarily for Team Portability (#2257) allowing developers to use different tools (Claude, Cursor, etc.) on the same repo without constant re-installs.

To avoid 'muddying the waters,' we use Logical Isolation:

Directory Silos: Agents look for their own folders (e.g., .claude/ vs .cursor/) and don't interfere with each other similar to having both .vscode and .idea folders.

State Switch: The default_integration metadata acts as a source of truth to identify the active agent.

If needed, we can enhance the use command to automatically update .gitignore or .cursorignore to hide inactive agents from the active one's context.

Does this isolation strategy work for you?

[mnriem]

You are saying they do not interfere, but you cannot guarantee that as LLMs will do what they do. So in order for me to accept this PR I request you to do the following:

1. Only allow installing more than one integration using the --force. If attempted without it indicate that installing multiple integration is not recommended
2. Merge list-integrations with list command
3. Update docs/reference/integrations.md to reflect the change. Also indicating it is NOT recommended