Use archive mode for nodes

I've found that if I wait too long between Smart Rollup origination and starting the Smart Rollup node, the node fails to start with this error:

Aug 01 15:41:22.865: Starting the smart rollup node
Aug 01 15:41:22.866: [sc_rollup_node] Starting layer 1 tracker
Aug 01 15:41:22.884: [Warning]: DAL is enabled in the protocol but no DAL node was provided for
Aug 01 15:41:22.884:   the rollup node
Aug 01 15:41:22.903: Starting refutation coordinator for the smart rollup node
Aug 01 15:41:22.905: The smart rollup node is listening to 127.0.0.1:8932
Aug 01 15:41:22.931: Processing head BKtVqgTJEHEAuhMNxSuoUiGMhupDhNWAVZMKHxduYMie6J7CcnX at level
Aug 01 15:41:22.931:   227
Aug 01 15:41:22.934: [Fatal daemon error]:
Aug 01 15:41:22.934:   Error:
Aug 01 15:41:22.934:     Did not find service: GET http://localhost:8732/chains/main/blocks/227/context/smart_rollups/all/inbox
Aug 01 15:41:22.934:
Aug 01 15:41:22.934: Restarting main rollup node loop in 10s.
Aug 01 15:41:23.321: [DEGRADED MODE] Seen layer 1 head
Aug 01 15:41:23.321:   BM58svpXYpM8Rf8qD1GZr8EWHq1sGZQnD3ax3uZRmyDWfRuwk4G at level 253
Aug 01 15:41:24.666: [DEGRADED MODE] Seen layer 1 head
Aug 01 15:41:24.666:   BLhrRecwDGqKzpamARxvXUNBbzh7Vv7BsgKXfTg2zmQF83JeQYn at level 254
^C(/usr/local/bin/octez-smart-rollup-node) INT: triggering shutdown.
Error:
  Lwt.Resolution_loop.Canceled

This may come about when the L1 node doesn't have enough history to satisfy the L2 node's needs when catching up to HEAD.

I don't seem to have this problem when the node is running in archive mode, so this MR changes it to archive mode in the sandbox setup.