Add spend analytics dashboard, improve session stopping, and refresh product docs/screenshots

[akhilsinghcodes]

This PR expands Agents Fleet beyond session launch/monitoring into a fuller cost visibility and control experience, while also tightening session lifecycle handling and updating the project docs/media to match the current product surface.

What’s included

1) Spend analytics dashboard

Adds a new Spend Dashboard UI that surfaces budget/cost visibility across several dimensions:

• Default overview
• Today
• 7 days
• By repo
• By command
• By model

This gives a much clearer view into where usage and cost are coming from, and helps users spot expensive workflows or repos faster.

2) Server-side dashboard support

Introduces backend support for the dashboard via a new server route:

• aggregates session/budget data
• exposes spend analytics to the web UI
• supports the different dashboard breakdowns shown in the new screenshots

3) Session lifecycle and timeout handling

Improves session management in the process manager and app/server flow:

• better handling for chat session timeouts
• improved stopping behavior
• cleaner lifecycle management for active sessions

4) UI updates

Updates the web app to support the new dashboard experience and related navigation/state changes.

5) Documentation refresh

Updates documentation to reflect the current product direction:

• README refreshed with new screenshots and visuals
• Architecture doc updated to mention:
  • PTY sessions
  • Claude SDK chat
  • LiteLLM-backed chat
  • persisted terminal replay
• Roadmap adjusted to reflect newly completed work

6) New assets

Adds a set of new visual assets for the spend dashboard and a new demo video:

• Spend dashboard screenshots
• Mission control video asset

Why this matters

This branch makes Agents Fleet more than just a session launcher/replayer. It now gives users:

• more visibility into spend
• better control over sessions
• clearer documentation of the current feature set

That’s a meaningful step toward the “mission control for AI agents” positioning.