Agent starting in semi-working state after unhandled promise rejection

Reports show that undetected errors in offchain subgraph ID parsing from environment variables cause the Agent to start in a partially operational mode.

This is the error log of an unhandled promise rejection involving a malformed deployment ID:

[2023-09-20 19:05:15.592 +0000] WARN (IndexerAgent/1 on d456316fe743): Unhandled promise rejection
    err: {
      "type": "IndexerError",
      "message": "Unhandled promise rejection",
      "stack":
          IndexerError: Unhandled promise rejection
              at indexerError (/opt/indexer/packages/indexer-common/dist/errors.js:173:12)
              at process.<anonymous> (/opt/indexer/packages/indexer-agent/dist/commands/start.js:351:56)
              at process.emit (node:events:527:28)
              at process.emit (node:domain:475:12)
              at emit (node:internal/process/promises:140:20)
              at processPromiseRejections (node:internal/process/promises:274:27)
              at processTicksAndRejections (node:internal/process/task_queues:97:32)
      "code": "IE035",
      "explanation": "https://github.com/graphprotocol/indexer/blob/main/docs/errors.md#ie035",
      "cause": {
        "type": "Error",
        "message": "Invalid subgraph deployment ID: QmQjmWt1vLf9aZmbpE8Amoyt5zJoawAJpEqwWJLAwXfkr8QmWfvBwdLL6e65Zddr6zbkcUY1mkyCKxyMLH33zTyfbJLn",
        "stack":
            Error: Invalid subgraph deployment ID: QmQjmWt1vLf9aZmbpE8Amoyt5zJoawAJpEqwWJLAwXfkr8QmWfvBwdLL6e65Zddr6zbkcUY1mkyCKxyMLH33zTyfbJLn
                at new SubgraphDeploymentID (/opt/indexer/node_modules/@graphprotocol/common-ts/dist/subgraphs/index.js:37:19)
                at /opt/indexer/packages/indexer-agent/dist/commands/start.js:500:149
                at Array.map (<anonymous>)
                at /opt/indexer/packages/indexer-agent/dist/commands/start.js:500:138
                at Generator.next (<anonymous>)
                at fulfilled (/opt/indexer/packages/indexer-agent/dist/commands/start.js:5:58)
                at processTicksAndRejections (node:internal/process/task_queues:96:5)
      }
    }

After that, only the monitorEthBalance, monitorNetworkPauses and monitorIsOperator routines were actively running.

Start with the startup path in packages/indexer-agent/dist/commands/start.js, especially the locations shown in the stack trace, and inspect how environment-provided subgraph deployment IDs are handled. Reproduce the malformed-ID case and trace the resulting promise rejection. Done should mean invalid configuration is handled explicitly rather than leaving the Agent partially operational.