fix(heartbeat): prevent false process_lost failures on queued and non-child-process runs

- reapOrphanedRuns() now only scans running runs; queued runs are
  legitimately absent from runningProcesses (waiting on concurrency
  limits or issue locks) so including them caused false process_lost
  failures (closes #90)
- Add module-level activeRunExecutions set so non-child-process adapters
  (http, openclaw) are protected from the reaper during execution
- Add resumeQueuedRuns() to restart persisted queued runs after a server
  restart, called at startup and each periodic tick
- Add outer catch in executeRun() so setup failures (ensureRuntimeState,
  resolveWorkspaceForRun, etc.) are recorded as failed runs instead of
  leaving them stuck in running state
- Guard resumeQueuedRuns() against paused/terminated/pending_approval agents
- Increase opencode models discovery timeout from 20s to 45s

packages/adapters/opencode-local/src/server/models.ts

@@ -7,6 +7,7 @@ import {
 } from "@paperclipai/adapter-utils/server-utils";
 
 const MODELS_CACHE_TTL_MS = 60_000;
+const MODELS_DISCOVERY_TIMEOUT_MS = 20_000;
 
 function resolveOpenCodeCommand(input: unknown): string {
   const envOverride =
@@ -115,14 +116,14 @@ export async function discoverOpenCodeModels(input: {
     {
       cwd,
       env: runtimeEnv,
-      timeoutSec: 20,
+      timeoutSec: MODELS_DISCOVERY_TIMEOUT_MS / 1000,
       graceSec: 3,
       onLog: async () => {},
     },
   );
 
   if (result.timedOut) {
-    throw new Error("`opencode models` timed out.");
+    throw new Error(`\`opencode models\` timed out after ${MODELS_DISCOVERY_TIMEOUT_MS / 1000}s.`);
   }
   if ((result.exitCode ?? 1) !== 0) {
     const detail = firstNonEmptyLine(result.stderr) || firstNonEmptyLine(result.stdout);