We can remove the known_query_shape_computation_problem tag from tests that run rawData commands once we have viewless time-series.

Context: 

Currently when running rawData commands on timeseries collections, the execution namespace will be switched to the system.buckets namespace. When looking up query settings, we skip generating the query shape hash if its a bucket namespace.

Test suites such as sharded_collections_query_shape_hash_stability check the query shape hash on potentially unsharded collections. On a shard that does not have the collection, we will not be able to translate the execution namespace to the buckets namespace (because it does not exist on this shard). Shards that do not have the timeseries collection will generate the query shape hash while shards with the collection will lookup query settings on the buckets collection and skip query shape generation. This results in failing this check that the query shape hash is identical on all nodes.

Once we have viewless time-series, there will no longer be a query shape hash discrepancy between shards with and without the collection.