GCP Pub Scaler should not panic

[nachocano]

GCP PubSub scaler panicking upon unmarshalling wrong credentials.

Expected Behavior

When passing invalid credentials to the gcp pub sub scaler, upon creation of the StackDriverClient, the code panics if the credentials cannot be unmarshalled.
See https://github.com/kedacore/keda/blob/master/pkg/scalers/stackdriver_client.go#L29

Actual Behavior

Log an error and return it.

Steps to Reproduce the Problem

1. Create a ScaledObject with PubSub scaler, with invalid credentials in the trigger metadata...
2. Publish messages to the topic in question.
3. The keda-operator pod will start crash looping

Specifications

• KEDA Version: master
• Platform & Version: GKE
• Kubernetes Version: 1.15
• Scaler(s): gcp pub sub

[tomkerkhove]

Can you let us know if it's still the case with 1.4.1 please?