Avoid panic by cheking whether entry is a span in batchCollector

[stoewer]

What this PR does:
It was observed that queries such as {} | rate() by (event.name) can cause panics in the batchCollector. Here is a stack trace taken from the logs:

2024-07-04 14:01:01.425	
github.com/grafana/tempo/pkg/traceql.(*MetricsEvalulator).Do(0xc0bd65db00, {0x3305e18, 0xc08a0a7440}, {0x32e1c60?, 0xc06f4f8cd0?}, 0x0?, 0xc099578f68?)
2024-07-04 14:01:01.425	
	/drone/src/tempodb/encoding/vparquet4/block_traceql.go:1312 +0x1f
2024-07-04 14:01:01.425	
github.com/grafana/tempo/tempodb/encoding/vparquet4.(*spansetIterator).Next(0xc097d151e0?, {0x3324f20?, 0xc08e5dad20?})
2024-07-04 14:01:01.425	
	/drone/src/tempodb/encoding/vparquet4/block_traceql.go:1198 +0x11f
2024-07-04 14:01:01.425	
github.com/grafana/tempo/tempodb/encoding/vparquet4.(*rebatchIterator).Next(0xc08a0a7f50)
2024-07-04 14:01:01.425	
	/drone/src/pkg/parquetquery/iters.go:1666 +0xa5
2024-07-04 14:01:01.425	
github.com/grafana/tempo/pkg/parquetquery.(*JoinIterator).Next(0xc04e206240)
2024-07-04 14:01:01.425	
	/drone/src/pkg/parquetquery/iters.go:1741 +0x352
2024-07-04 14:01:01.425	
github.com/grafana/tempo/pkg/parquetquery.(*JoinIterator).collect(0xc04e206240, {0xc9, 0x0, 0x0, 0x0, 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff})
2024-07-04 14:01:01.425	
	/drone/src/pkg/parquetquery/iters.go:1863 +0xd6
2024-07-04 14:01:01.425	
github.com/grafana/tempo/pkg/parquetquery.(*LeftJoinIterator).Next(0xc0631158c0)
2024-07-04 14:01:01.425	
	/drone/src/tempodb/encoding/vparquet4/block_traceql.go:2707 +0xcb1
2024-07-04 14:01:01.425	
github.com/grafana/tempo/tempodb/encoding/vparquet4.(*batchCollector).KeepGroup(0xc08a0a7ef0, 0xc0c47cc960)
2024-07-04 14:01:01.425	
goroutine 62344 [running]:
2024-07-04 14:01:01.425	

2024-07-04 14:01:01.425	
panic: interface conversion: interface {} is traceql.Static, not *vparquet4.span

I wasn't able to reproduce this locally or in tests, but given the above log the missing check when casting to *span looks like the most likely culprit.

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]