Name server threads for Linux diagnostics

Server threads currently appear under the same process name, making it difficult to identify which thread is accepting connections or consuming CPU in service, fabric, storage, and background work. This change assigns short role-based names that appear in Linux thread listings and profilers. Service workers, fabric senders, and SSD writers include their existing worker indices.

The shared thread wrapper copies each name before a new thread or pooled task starts and sets it from within the running thread. Reusable pool threads take the submitted task's name while working and return to as-pool-idle after the task and its cleanup. Existing thread creation APIs remain available.

This lets operators identify the acceptor's TID and investigate listen-queue overflows alongside its CPU use or blocking behavior. The change affects observability; it does not alter the accept backlog, scheduling, affinity, or transaction handling.

Validation: git diff --check passed, and a source check confirmed named calls have the expected arguments and literal names fit Linux's 15-byte limit. A Linux build and server smoke test remain outstanding because the available checkout is on macOS with uninitialized submodules.