Fix #542: correct the couchbase-source statement example

[oscerd]

Fixing a defect I introduced in #3007.

The statement example shipped there does not work:

        example: "SELECT * FROM `travel-sample` LIMIT 10"

and it fails in the worst possible way — the source polls happily, logs nothing alarming, and emits no exchanges at all.

Two requirements, neither obvious

Both found by running the Kamelet against a live Couchbase while writing the integration test in #3008.

1. The document id must be aliased as __id. CouchbaseConsumer does:

String id = row.getString(SQL_DOCUMENT_ID_ALIAS);   // "__id"
if (id == null) {
    LOG.warn("Row does not contain '{}' field. ... Skipping row.", ...);
    continue;
}

A query without META().id AS __id has every row skipped — at WARN level, so a route looks healthy and simply never fires.

2. The keyspace is the collection, not the bucket. The consumer runs the query through scope.query(...), so FROM <bucket> resolves to <bucket>._default.<bucket>:

Keyspace not found in CB datastore: default:mybucket._default.mybucket

The collection — normally _default — is what belongs there.

The change

-        description: The N1QL query to run against the bucket on each poll. Used unless useView is true.
+        description: >-
+          The N1QL query to run against the bucket on each poll. Used unless useView is true.
+          The query runs in the bucket scope, so the keyspace is the collection - use _default
+          for the default collection rather than the bucket name. It must also alias the document
+          id as __id, because rows without that field are skipped.
-        example: "SELECT * FROM `travel-sample` LIMIT 10"
+        example: "SELECT META().id AS __id, * FROM _default"

Description only — no template or behavioural change. The example now matches exactly what the integration test in #3008 uses and is proven to work against a real cluster.

Kept separate from #3008 deliberately: that PR adds a test, this one corrects a Kamelet, and they are reviewable independently even though the same investigation produced both.

script/validator reports no errors and mvn clean install passes with tests from the repository root.

---

Claude Code on behalf of Andrea Cosentino