Support window functions in time-series aggregations (#138139)

This change adds window function support for time-series aggregation 
functions including `rate`, `avg_over_time`, and `last_over_time`.
Additional time-series aggregation functions will be supported in a
follow-up to keep this PR focused.


**Examples:**

```js
TS k8s
| WHERE TRANGE("2024-05-10T00:05:00.000Z", "2024-05-10T00:10:00.000Z")
| STATS rate_bytes_in=avg(rate(network.total_bytes_in, 5minute)) 
              BY cluster, time_bucket = bucket(@timestamp, 1minute)
```

Window 5-minute over 1-minute buckets.

```js
TS k8s 
| STATS events = sum(avg_over_time(events_received, 10minute)) 
             BY pod, time_bucket = tbucket(5minute)
| SORT time_bucket, pod
| LIMIT 10
```

Window 10-minute over 5-minute buckets.

**Limitation**: 

ES|QL currently supports only windows that are multiples of the bucket size.

Author: dnhatn

docs/changelog/138139.yaml

@@ -0,0 +1,5 @@
+pr: 138139
+summary: Support window function in time-series aggregations
+area: ES|QL
+type: enhancement
+issues: []