Too many ThreadLocals in Hedera node

### Problem

`jmap` histogram of a Hedera node shows 1700+ `ThreadLocals` most of which come from `log4j`. There is performance impact on operations with Hedera's own `ThreadLocals`.

The reason is a misuse of `ThreadLocal` due to not understanding the basic patterns of its use: a new `ThreadLocal` is created for each instance of `ReusableMessageFactory` (i.e. for each `Logger`), `AppenderControl`, etc.
See f.e. https://github.com/apache/logging-log4j2/commit/e86d28bdb927f8047b924eb459f4c63bd2e43176

### Solution

- file a ticket against `log4j` explaining the pattern of `static ThreadLocal.withInitial()`

### Alternatives

_No response_

Start by reviewing the jmap histogram findings in the issue and the linked Log4j commit e86d28bdb927f8047b924eb459f4c63bd2e43176. Compare the described per-instance ThreadLocal pattern with static ThreadLocal.withInitial(), then file the requested Log4j ticket explaining the issue and its performance impact.

Written by the indexing model from the issue text.