Allow executing mutations / subscriptions in parallel / batched mode

[kyri-petrou]

Came across this while optimizing the Executor. Currently we force the execution of mutations and subscriptions to sequential mode

caliban/core/src/main/scala/caliban/execution/Executor.scala

Lines 43 to 47 in 078c28d

	val execution = request.operationType match {
	case OperationType.Query => queryExecution
	case OperationType.Mutation => QueryExecution.Sequential
	case OperationType.Subscription => QueryExecution.Sequential
	}

According to the spec, only the top-level fields are required to be executed sequentially (at least for mutations, couldn't find anything about subscriptions requiring them to be executed serially). Ideally, we'd want to execute the top-level fields of mutations sequentially, but execute any child fields using the user-provided execution mode.