[Dashboard] Align run retention documentation with implementation

### Is there an existing issue for this?

- [x] I have searched the existing issues

### Describe the bug

The Dashboard persistence specification and implementation describe different retention contracts:

- One section says `Run` retains five unpinned historical directories and current does not count.
- Another says `Run` retains ten databases.
- The PR description says up to ten runs per application.
- Code retains ten unpinned runs including current, which means current plus nine unpinned historical runs, in addition to pinned runs.

### Expected Behavior

The implementation, tests, PR description, and specification should describe one consistent retention policy.

### Steps To Reproduce

1. Read `docs/specs/dashboard-persistence.md` around lines 163-165 and 298-300.
2. Compare with `DashboardRunStore.MaxRuns` and `PruneRuns`.
3. The documented counts and whether current consumes the budget disagree.

### Exceptions (if any)

N/A

### Aspire doctor output

```plain
Aspire CLI 13.6.0-pr.18924.g5ee63941
```

### Anything else?

Related to #18924. Source: `docs/specs/dashboard-persistence.md:163-165,298-300` and `src/Aspire.Dashboard/ServiceClient/DashboardRunStore.cs:74,372-376,440-442`.

Start with docs/specs/dashboard-persistence.md at lines 163-165 and 298-300, then compare the stated policy with DashboardRunStore.MaxRuns and PruneRuns in src/Aspire.Dashboard/ServiceClient/DashboardRunStore.cs at the cited lines. Confirm the implementation's treatment of current, unpinned, historical, and pinned runs, then make the documented retention contract consistent with the implementation, tests, and related PR description.

Written by the indexing model from the issue text.