fix(clickhouse): add execution time cap to task run metrics queries

`getTaskActivity` and `getAverageDurations` were registered without any
`clickhouseSettings`, so neither had a server-side `max_execution_time`.
Their sibling `existsQueryBuilder` in the same object literal already
passes one.

With no server cap, the only thing bounding a slow query is the client's
`request_timeout`. That aborts the HTTP request rather than the query, so
the failure surfaces as an untyped socket timeout instead of a ClickHouse
timeout error, and stopping the query itself depends on the server
noticing the disconnect.

Set `max_execution_time: 25` on both. `@clickhouse/client` defaults
`request_timeout` to 30000ms and nothing in this repo overrides it, so 25s
sits below the client timeout — ClickHouse terminates the query and
returns a typed error before the client gives up. Mirroring the
neighbouring `10` would instead start failing queries that currently
succeed in the 10-30s range.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PtQ34usZSzrSALPVBmoLYA

Author: claude

.server-changes/task-run-metrics-query-time-limit.md

@@ -0,0 +1,6 @@
+---
+area: webapp
+type: fix
+---
+
+Task metrics that take too long to load now stop with a clear error instead of hanging until the request gives up

internal-packages/clickhouse/src/index.ts

@@ -232,10 +232,16 @@ export class ClickHouse {
       existsQueryBuilder: getTaskRunExistsQueryBuilder(this.reader, { max_execution_time: 10 }),
       tagQueryBuilder: getTaskRunTagsQueryBuilder(this.reader),
       pendingVersionIdsQueryBuilder: getPendingVersionIdsQueryBuilder(this.reader),
-      getTaskActivity: getTaskActivityQueryBuilder(this.reader),
+      // Cap these server-side. The client's `request_timeout` defaults to 30s, and
+      // without a `max_execution_time` that timeout is the only thing that stops a
+      // slow query — it aborts the HTTP request rather than the query, so the
+      // failure arrives as an untyped socket timeout. 25s keeps every query that
+      // currently succeeds while letting ClickHouse terminate its own work first
+      // and return a proper timeout error.
+      getTaskActivity: getTaskActivityQueryBuilder(this.reader, { max_execution_time: 25 }),
       getCurrentRunningStats: getCurrentRunningStats(this.reader),
       getChildRunStatusCounts: getChildRunStatusCounts(this.reader),
-      getAverageDurations: getAverageDurations(this.reader),
+      getAverageDurations: getAverageDurations(this.reader, { max_execution_time: 25 }),
       getTaskUsageByOrganization: getTaskUsageByOrganization(this.reader),
     };
   }