[ESQL] Log compute engine server failures as ERROR

[MattAlp]

Closes #139883.

Keep 4xx driver failures at DEBUG, and log non-4xx runtime driver failures at ERROR (was WARN).
Add DriverTests coverage to verify server-side failures log at ERROR, while 4xx failures do not log at WARN/ERROR.

Validated with:
./gradlew :x-pack:plugin:esql:compute:test --tests "org.elasticsearch.compute.operator.DriverTests"

(TIL that gh + LLMs go well together)

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[elasticsearchmachine]

Hi @MattAlp, I've created a changelog YAML for you.

[alex-spies]

Thanks @MattAlp , I think this is the right solution.

Left some suggestions for expanded test coverage. It'd would be nice if you could look into the additional cases, but apart from that this is good to go from my side and doesn't require re-review from me.

[alex-spies]

Oh, and I also locally tried this by reproducing the current bug in FIRST/LAST, and it logs as expected, even when running async. Great :)

curl -u elastic:password -H "Content-Type: application/json" "127.0.0.1:9200/_query/async" -d '
{ "wait_for_completion_timeout": "1ms",
  "query": "row x = 1::long | stats y = first(x, x)"}'


[2026-02-12T11:07:56,530][ERROR][o.e.c.o.Driver           ] [runTask-0] Error running driver [single] java.lang.ArrayIndexOutOfBoundsException: Index 1 out of bounds for length 1