Expand MCP beyond coding harness (e.g. dashboard management via natural language)

## Summary

`sentry-mcp` is scoped today as a coding/debugging harness. This issue proposes expanding the MCP's surface to cover non-debugging Sentry capabilities, starting with dashboard management via natural language.

## Problem

The README explicitly calls out:

> "Our tool selection and priorities are focused on developer workflows and debugging use cases, rather than providing a general-purpose MCP server for all Sentry functionality."

That's a reasonable starting point, but it leaves a lot of value on the table. Users want to interact with Sentry's broader operational surface — dashboards, alerts, monitors — through AI agents, not just debug errors.

## Proposed Direction

1. Audit which Sentry API surfaces are high-value for LLM interaction but currently out of scope.
2. Add dashboard management as the first non-debugging expansion (create, list, update, delete dashboards and widgets).
3. Establish a pattern for non-debugging skills so future additions (alerts, monitors) are straightforward.
4. Update the README to reflect the expanded scope.

## Notes

- Sentry's REST API has full CRUD for dashboards (`/api/0/organizations/{org}/dashboards/`).
- The existing `--disable-skills` / `MCP_DISABLE_SKILLS` flag makes new skills opt-out-able.
- Token scopes for dashboard writes need to be verified.