`shard_collection_error_handling.js` validates the correct behavior of shardCollection command when finding errors and it is needed to rollback or continue the operation. In order to do that the test triggers the primary node to stepdown and validate how the coordinator resumes in the new primary.

When this test is run on Windows machines, the election of the new primary can be slow and make validations timed out.

I propose to ban this test from windows machines as it is not giving us more coverage to run it in slow variants as we only want to test this in specific scenarios.

No need to backport this ticket.