fix(server): decouple asyncify thread pool from COOP_TASKRUN for old-kernel compat

[mfyuce]

Summary

IORING_SETUP_COOP_TASKRUN requires Linux ≥ 5.19. On kernels 5.10–5.18 the
shard io_uring setup fails with EINVAL even though the main runtime starts
fine, preventing server boot entirely. These three commits fix that in order:

1. Gate COOP_TASKRUN flags behind an env var (IGGY_SHARD_RUNTIME_COOP_TASKRUN,
   default true = unchanged). Set it to false to run on 5.10–5.19 kernels
   at a small latency cost. (feat(server))

2. Keep asyncify worker pool when COOP_TASKRUN is off. With the flag off,
   compio routes fs/JWT-storage ops through the asyncify thread pool.
   thread_pool_limit(0) then panics "thread pool is needed but no worker
   thread is running" on shard 0. Gate the thread_pool_limit(0) call behind
   the same flag. (fix(server))

3. Decouple thread_pool_limit from COOP_TASKRUN entirely. TCP, HTTP, and
   WebSocket transports dispatch some ops through the asyncify pool even when
   COOP_TASKRUN=true, so tying thread_pool_limit(0) to the flag still
   panics on 6.8+ kernels with those transports active. Add a
   keep_worker_pool: bool parameter to create_shard_executor; the pool is
   only dropped when COOP_TASKRUN=true and no TCP/HTTP/WS transport is
   enabled. Both server and server-ng derive keep_worker_pool from their
   loaded config. (fix(server))

After these changes, operators can set IGGY_SHARD_RUNTIME_COOP_TASKRUN=true
on Linux 6.8+ with TCP transport enabled and get lower io_uring latency without
the worker-pool panic. On ≤5.18 kernels they set it to false and the server
boots normally.

Files changed

• core/server_common/src/executor.rs — create_shard_executor(keep_worker_pool: bool), flag-gated COOP_TASKRUN/TASKRUN_FLAG
• core/server/src/main.rs — derive keep_worker_pool from config
• core/server-ng/src/main.rs, core/server-ng/src/bootstrap.rs — same

Test plan

• cargo clippy -p server -p server-ng -- -D warnings passes
• Server boots on Linux 5.15 with IGGY_SHARD_RUNTIME_COOP_TASKRUN=false + TCP transport
• Server boots on Linux 6.8 with IGGY_SHARD_RUNTIME_COOP_TASKRUN=true + TCP transport (no worker-pool panic)
• cargo test -p server_common passes

🤖 Generated with Claude Code

https://claude.ai/code/session_018duZYBkbguQ2pn8RJ82PUw

[github-actions]

Thanks for the PR. It is labeled S-waiting-on-review and queued for review.

Slash commands (own line, regular comment) move it around the queue:

• /ready - back to S-waiting-on-review after addressing feedback
• /author - flip to S-waiting-on-author while you finish changes
• /request-review @user-or-team - request a reviewer

See CONTRIBUTING.md for details.

[numinnex]

I am not sure about it, I think we rather not allow our server to run on kernels < 6.8, rather than adding extra startup flags that to disable COOP_TASKRUN.