create dedicated filesystem check for GKE nodes, and remove from predict_linear

Removes GKE nodes from the current filesystem check, and adds a new dedicated one with a static threshold.
GKE node storage space can fluctuate and should generally be handled by the kubernetes garbage collection.
Because of this the predict_linear filesystem check can lead to excessive noise.
By using a dedicated static threshold of 90% we only alert if it has gone past the kuberenets garbage collection threshold, which would generally indicate we legitimately are running out of disk space on the node.