fix: Ensure Flux reads across all shards

[stuartcarnie]

Closes #20035

This PR ensures all elements of the cursors.CursorIterator slice (itrs) are read when reading a multi-shard cursor. It is accompanied with a unit test to replicate the incorrect behaviour.

The root cause is that it is possible a single shard has no data for a given query and this resulted in the read stopping short, as it returned a nil cursor.

Unlike the createCursor function, which finds the next non-nil result from the itrs slice:

influxdb/storage/reads/array_cursor.go

Lines 110 to 113 in b7ac9f0

	for cur == nil && len(row.Query) > 0 {
	shard, row.Query = row.Query[0], row.Query[1:]
	cur, _ = shard.Next(m.ctx, &m.req)
	}

the existing implementation of nextArrayCursor was only checking one element of the itrs slice:

influxdb/storage/reads/array_cursor.gen.go

Lines 316 to 317 in 632e8f1

	itr, c.itrs = c.itrs[0], c.itrs[1:]
	cur, _ = itr.Next(c.ctx, c.req)

If the call to Next returned a nil Cursor, that signalled the end of the read.

This fix ensures itrs is read in a loop for the next non-nil cursor.

• CHANGELOG.md updated with a link to the PR (not the Issue)
• Well-formatted commit messages
• Rebased/mergeable
• Tests pass

[stuartcarnie]

@nathanielc please feel free to add someone from your team for review.

@danxmoran I added you, so you could see what the fix was 👍