only-L1 node fall into infinite loop with `SkipSyncStartCheck == true`

**Bug Description**
node with state `unsafe == safe != 0` will run into infinite loop during `FindL2Heads` if it restarts with `SkipSyncStartCheck == true`.

**Steps to Reproduce**
Simply spin up a node without any p2p settings to let it only derive blocks from L1. Once it derived some blocks (unsafe != 0), restart it with env `OP_NODE_L2_SKIP_SYNC_START_CHECK: true`. That is it. Then node will print this log infinitely `Found highest L2 block with canonical L1 origin. Skip further sanity check and jump to the safe head`

**Expected behavior**
skip walking back unsafe blocks(in this case, there is no unsafe blocks) and continue for safe blocks until finding a safe block of which sequence number is 0 and the belonging epoch has been fully rolled up(sequencing window passed).

**Environment Information:**
op-node v1.16.3

**Configurations:**
no p2p settings and restart with `OP_NODE_L2_SKIP_SYNC_START_CHECK: true`

**Logs:**
infinite `Found highest L2 block with canonical L1 origin. Skip further sanity check and jump to the safe head`

Start at the FindL2Heads entry point and reproduce the restart scenario with no P2P settings and OP_NODE_L2_SKIP_SYNC_START_CHECK set to true. Trace the repeated “Found highest L2 block with canonical L1 origin” path and verify that the node skips unsafe-block walking, continues through safe blocks, and stops once it finds a fully rolled-up epoch with sequence number 0.

Written by the indexing model from the issue text.