Date_histogram doesn't work too well with timezones

[trzysiek]

We'll need to think how to tackle time zones, in both Elastic, and Clickhouse.
How to reproduce:
I changed docker/log-generator/logger's code for timestamp to:

stamps := []time.Time{
  time.Date(2021, 1, 2, 0, 0, 0, 0, time.UTC),
  time.Date(2021, 1, 1, 23, 50, 0, 0, time.UTC),
  time.Date(2021, 1, 1, 22, 55, 0, 0, time.UTC),
  time.Date(2021, 1, 2, 0, 55, 0, 0, time.UTC),
  time.Date(2021, 1, 2, 1, 55, 0, 0, time.UTC),
}
...
"@timestamp":   stamps[rand.Intn(len(stamps))].Format("2006-01-02T15:04:05.999Z"),

Then, there's discrepancy between our and Elastic's responses for query at the bottom. Ratio between hits in those 2 buckets is very different.
Elastic returns

And we do:

Query:

POST /logs-generic-default/_search
{
    "_source": {
        "excludes": []
    },
    "aggs": {
        "0": {
            "date_histogram": {
                "fixed_interval": "1d",
                "field": "@timestamp",
                "min_doc_count": 1,
                "time_zone": "Europe/Warsaw"
            }
        }
    },
    "fields": [
        {
            "field": "@timestamp",
            "format": "date_time"
        }
    ],
    "query": {
        "bool": {
            "filter": [
                {
                    "range": {
                        "@timestamp": {
                            "format": "strict_date_optional_time",
                            "gte": "2020-03-11T04:08:33.783Z",
                            "lte": "2024-06-11T03:08:33.783Z"
                        }
                    }
                }
            ],
            "must": [],
            "must_not": [],
            "should": []
        }
    },
    "runtime_mappings": {},
    "script_fields": {},
    "size": 0,
    "stored_fields": [
        "*"
    ],
    "track_total_hits": true
}

[jakozaur]

Fixed.