fix(remote-ssh): use remote user's real shell, keep dead panes alive

Remote shell-mode sessions hardcoded 'exec bash -l', ignoring the
remote user's actual login shell. sshd sets $SHELL from the remote
user's /etc/passwd entry, so 'exec $SHELL -i -l' launches their real
shell (zsh, fish, etc.) with rc files sourced, same fix as the local
shell-mode launch.

Also set remain-on-exit on the remote tmux session. It was only ever
set on the local socket, so if the remote command exited for any
reason -- even something transient -- tmux destroyed the pane, window,
and (being the only session) the whole remote server, tearing down the
local ssh attach along with it and leaving no trace to diagnose. The
local pane saw this as an instant clean exit, and reconnect's -A then
created a fresh session, which could repeat as a flap loop with no
evidence surviving between attempts.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

src/remote-hosts.ts

@@ -60,7 +60,11 @@ export async function writeRemoteCases(configDir: string, cases: RemoteCase[]):
 
 export function defaultRemoteCommandForMode(mode: SessionMode): string {
   const commands: Record<RemoteCommandMode, string> = {
-    shell: 'exec bash -l',
+    // $SHELL, not a hardcoded bash: sshd sets it from the remote user's
+    // /etc/passwd entry, so this launches their actual login shell (zsh,
+    // fish, etc.). -i -l so it sources rc files (~/.zshrc etc.), matching
+    // the local shell-mode launch.
+    shell: 'exec $SHELL -i -l',
     // Mirror the LOCAL claude default so the remote agent runs non-interactively
     // (no trust-folder/permission prompt that nothing on the remote answers). The
     // per-host `commands.claude` override stays the escape hatch.