cloud: retry a failed key_cmd on the next models requestHEADmaster

Keys were read once at startup. At login the smartcard PIN is often not
entered within key_cmd's timeout, so that provider was left out for the
whole run and its models stayed missing until a restart.

send_models now asks again for every provider without a key. It runs off
the lock on purpose: route() holds it and must not wait on a PIN prompt.
The refusal message says to open the model list to retry, and the
self-check covers a key that fails at startup and loads on the next
models request.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

0 files changed, 0 insertions, 0 deletions