evaluation error on CB class test expressions

1. Start a Kea with CB and debug logging kea-dhcp4.conf.
2. Start a kea-ctrl-agent kea-ctrl-agent.conf.
3. Create a class that tests the equality operator between an option substring and a literal string:

$ curl -X POST 10.1.0.2:8080 -H 'Content-Type: application/json' -d "$(cat <<HERE_DOCUMENT
{
  "arguments": {
    "client-classes": [
      {
        "name": "my-class",
        "test": "substring(option[1].hex, 0, 8) == 'my-value'"
      }
    ],
    "remote": {
      "type": "mysql"
    },
    "server-tags": [
      "all"
    ]
  },
  "command": "remote-class4-set",
  "service": [
    "dhcp4"
  ]
}
HERE_DOCUMENT
)"
[ { "arguments": { "client-classes": [ { "name": "my-class" } ] }, "result": 0, "text": "DHCPv4 client class successfully set." } ]

4. Wait for Kea to pull configuration signaled through the log message: DEBUG MYSQL_CB_GET_ALL_CLIENT_CLASSES4_RESULT retrieving: 1 elements
5. Send perfdhcp traffic: perfdhcp -4 -l vethclient -r 1 -t 1.
6. Observe error log message: ERROR EVAL_RESULT Expression my-class evaluated to Incorrect stack order. Expected exactly 1 value at the end of evaluation, got 0.

Expected result: the my-class expression gets succesfully evaluated to either true or false.

Alternative scenario: configuring the client class in the file configuration evalutes the expression succesfully to true or false.

"client-classes": [
  {
    "name": "my-class",
    "test": "substring(option[1].hex, 0, 8) == 'my-value'"
  }
]

DEBUG EVAL_DEBUG_OPTION Pushing option 1 with value 0x
DEBUG EVAL_DEBUG_STRING Pushing text string '0'
DEBUG EVAL_DEBUG_STRING Pushing text string '8'
DEBUG EVAL_DEBUG_SUBSTRING_EMPTY Popping length 8, start 0, string 0x pushing result 0x
DEBUG EVAL_DEBUG_STRING Pushing text string 'my-value'
DEBUG EVAL_DEBUG_EQUAL Popping 0x6D792D76616C7565 and 0x pushing result 'false'
DEBUG EVAL_RESULT Expression my-class evaluated to 0

Hint: in CB's case, EVAL_DEBUG_* messages are missing from the logs. I would suggest logging them in this MR if the estimated effort is minimal.

RT #18874