perf: save and exploit distrinction between resources and attributes

[ernado]

Following #286:

During query execution, we should first analyze all label matches and group them into "attribute-only", "resource-only" or "both".

To do it fast, there should be a materialized view or additional table that holds this information.

So, before:

SELECT name, timestamp, mapping, value, flags, attributes, resource
FROM `metrics_points`
WHERE true
  AND toUnixTimestamp64Nano(timestamp) >= 1704546540000000000
  AND toUnixTimestamp64Nano(timestamp) <= 1704549765000000000
  AND (
    JSONExtractString(attributes, 'instance') = 'host-0'
        OR JSONExtractString(resource, 'instance') = 'host-0'
    )
  AND (
    JSONExtractString(attributes, 'job') = 'node_exporter'
        OR JSONExtractString(resource, 'job') = 'node_exporter'
    )
  AND (
    JSONExtractString(attributes, 'mode') = 'user'
        OR JSONExtractString(resource, 'mode') = 'user'
    )
  AND (
    name = 'node_cpu_seconds_total'
    )
ORDER BY timestamp;

After:

SELECT name, timestamp, mapping, value, flags, attributes, resource
FROM `metrics_points`
WHERE true
  AND toUnixTimestamp64Nano(timestamp) >= 1704546540000000000
  AND toUnixTimestamp64Nano(timestamp) <= 1704549765000000000
  AND JSONExtractString(resource, 'instance') = 'host-0'
  AND JSONExtractString(resource, 'job') = 'node_exporter'
  AND JSONExtractString(attributes, 'mode') = 'user'
  AND name = 'node_cpu_seconds_total'
ORDER BY timestamp;