feat: dynamic user idle_timeout from patch.cfg without container restart

[rippleitinnz]

Problem

user.idle_timeout is only read at startup in usr::init() and cached in
metric_thresholds[4]. It cannot be changed on a running node without a
container restart, which is not possible on leased Evernode instances.

Currently user.idle_timeout defaults to 0 (unlimited) in most deployments.
However operators may wish to set a timeout to protect against stale client
connections consuming resources, or to adjust it dynamically based on workload.
Without this fix, any change to user.idle_timeout requires a full container
restart to take effect.

Fix

Three changes:

comm_server.hpp — added for_each_session() template method to iterate
over all active sessions under mutex protection. Shared with sibling PR
feat/dynamic-mesh-idle-timeout-from-patch-cfg.

usr.cpp — added update_idle_timeout() which updates metric_thresholds[4]
for future connections AND calls set_threshold(IDLE_CONNECTION_TIMEOUT) on all
existing active user sessions via for_each_session().

conf.cpp — reads user.idle_timeout from patch.cfg in apply_patch_config(),
calls usr::update_idle_timeout() when value changes.

Effect

Operators can now update user.idle_timeout via patch.cfg on a running cluster
without any container restart. Takes effect immediately on all active and future
user connections.

Sibling PRs

This is part of a series making some hpcore config fields dynamically updatable from patch.cfg:

• fix/dynamic-log-level-from-patch-cfg — feat: dynamic log level update from patch.cfg without container restart #413
• feat/dynamic-mesh-idle-timeout-from-patch-cfg — feat: dynamic mesh idle_timeout from patch.cfg without container restart #414

These three PRs should be reviewed and merged together. This PR and #414 share
the for_each_session() template added to comm_server.hpp.