This reverts commit 8135224.

As reported in https://sourcegraph.slack.com/archives/C023ELQLV7F/p1672835804462349.

Currently, Zoekt tries to report in its responses whether or not it has finished the initial load of all of its shards before performing the query. However, there was a bug that caused to still report that it hadn't finished loading even if there were no shards to load. This bug caused this error message to appear on the site admin page:

This revert papers over this issue by restoring the old behavaior (not outright failing the ListAllIndexed call if a crash occurred during the response). After the revert, the site-admin page looks normal again:

The underlying cause in Zoekt was fixed in sourcegraph/zoekt#512. However, for the purposes of a 4.3.1 release I think this change is the smaller one to make to work around the issue (Zoekt has recieved other commits since 4.3.0 release, and I don't want to bring in more changes than we have to).

Test plan

Manual testing:

1. Run the local dev stack with the external service commented out:

   {
     // "GITHUB": [
     //   {
     //     "url": "https://github.com",
     //     "token": "$REDACTED", // user: sourcegraph-dogfood-user token: dev-private-localdev
     //     "repos": [
     //       "sourcegraph/sourcegraph",
     //       "hashicorp/go-multierror",
     //       "hashicorp/errwrap",
     //       "sourcegraph-testing/etcd",
     //       "sourcegraph-testing/tidb",
     //       "sourcegraph-testing/titan",
     //       "sourcegraph-testing/zap"
     //     ]
     //   }
     // ]
   }

2. Go to the site admin page, and see this screenshot:
   
3. Revert 8135224, and go the site admin page again and see that the page looks normal again: