plat: apl: clocks gated back on topology free

[mmaka1]

Signed-off-by: Marcin Maka marcin.maka@linux.intel.com

[mmaka1]

@lgirdwood This is implementation of the put-side for the dma and dai devices (see the #452).

While testing, I found that the ssp dai was not removed at the end as expected (3 refs obtained and 1 left). There was Tx dai comp, Rx dai comp, and the third unbalanced one appeared to be the IPC ipc_dai_config() handler. It is not a big deal to add the 'put' operation at the end of the handler but there is another thing. It might be more reliable to introduce 'CREATE' flag to dai_get() and dma_get(), so that the components specify CREATE meaning the object should be 'probed' if does not exist while the IPC config handlers should request only the existing objects and fail otherwise. It is to avoid a scenario, and fail early, where the driver implementation would incorrectly call (1) config first and then the (2) dai component new, so that the device would be (1) created-configured-destroyed, and then (2) created-and-used-unconfigured.