[FEA] Harden Java native-resource cleanup and close safety

**Describe the bug**

Several Java wrappers can leak retained resources if native destruction throws. Native operations can also race with `close()`, producing use-after-free, while calls after closure may throw incidental `NullPointerException`s instead of `IllegalStateException`.

**Steps/Code to reproduce bug**

Source-level examples include `HashJoin` and `KeyRemapping` cleanup, concurrent native use and closure, and accessing wrapper metadata after closure. Audit analogous multi-resource cleaners such as `HostColumnVectorCore`.

**Expected behavior**

- Always attempt all cleanup and preserve suppressed exceptions.
- Prevent native handles from being destroyed while in use.
- Explicitly reject use after closure.
- Consolidate common cleaner behavior/naming and simplify `cleanImpl` control flow where appropriate.

**Environment overview (please complete the following information)**

All environments; identified through source review.

**Environment details**

N/A.

**Additional context**

Follow-up to NVIDIA/cudf#22392, particularly the review discussions about [exception-safe cleanup](https://github.com/NVIDIA/cudf/pull/22392#discussion_r3907074093), [close/use races](https://github.com/NVIDIA/cudf/pull/22392#discussion_r3907241235), and [use after close](https://github.com/NVIDIA/cudf/pull/22392#discussion_r3908080644).

Start with the HashJoin and KeyRemapping cleanup examples, then audit the analogous multi-resource cleaner HostColumnVectorCore. Read the linked review discussions on exception-safe cleanup, close/use races, and use after close before deciding the scope. Done means cleanup attempts all resources with suppressed exceptions, native use is safe against close, and post-close access explicitly raises IllegalStateException.

Written by the indexing model from the issue text.