Add fleet-wide instance capability coverage matrix

### Problem

RocketMQ Studio already exposes an instance capability contract and uses it to hide unsupported navigation for the currently selected instance. Operators managing mixed Apache, Alibaba Cloud, and Tencent Cloud fleets still have no way to compare those declarations before planning an operation or rollout. They must enter instances one by one, and a discovery failure can easily be confused with an unsupported capability.

### Proposed change

Add a read-only capability matrix to the Instance page that:

- loads the existing server-declared capability contract for selected instances in bounded batches;
- preserves successful rows when individual instances fail and marks failures as unknown rather than unsupported;
- summarizes coverage for Topic, Consumer Group, message query, trace, ACL, and DLQ management;
- filters by vendor, access type, capability, support state, discovery state, and text;
- exports the currently filtered matrix to CSV;
- provides Chinese and English copy plus focused unit/component tests.

The feature does not infer capabilities, mutate instances, or call operation endpoints.

Start at the Instance page and trace how the existing server-declared capability contract is loaded and used to hide unsupported navigation. Define the matrix around bounded discovery, failure states, filtering, CSV export, and Chinese/English copy, then add focused unit/component tests covering these behaviors; done means the filtered read-only matrix preserves successful rows and distinguishes unknown from unsupported.

Written by the indexing model from the issue text.