Add guard against missing samples in KEDA queries

[jhalterman]

What this PR does

This PR adds clauses to some of our KEDA queries to guard against scenarios that could cause unintended scaledowns. This happens when the prometheus targeted by KEDA is down or unresponsive for some time, which could cause data that is used in KEDA queries to be missed. Without this data, when prometheus comes back up, these queries return lower averages for a while, which can cause unintended scaledowns.

An example of what this looks like from kubectl before, during, and after prometheus going down:

NAME                   REFERENCE                TARGETS                                             MINPODS   MAXPODS   REPLICAS   AGE
keda-hpa-distributor   Deployment/distributor   1046561m/1k (avg), 1633790418097m/2147483648 (avg)   10        40        29         171d
keda-hpa-distributor   Deployment/distributor   <unknown>/1k (avg), 0/2147483648 (avg)               10        40        29         171d
keda-hpa-distributor   Deployment/distributor   61190m/1k (avg), 488859859863m/2147483648 (avg)      10        40        29         171d
keda-hpa-distributor   Deployment/distributor   81245m/1k (avg), 488973756911m/2147483648 (avg)      10        40        26         171d
keda-hpa-distributor   Deployment/distributor   81245m/1k (avg), 488973756911m/2147483648 (avg)      10        40        26         171d
keda-hpa-distributor   Deployment/distributor   274241m/1k (avg), 1511713092856m/2147483648 (avg)   10        40        26         171d
keda-hpa-distributor   Deployment/distributor   275528m/1k (avg), 1511747993600m/2147483648 (avg)   10        40        26         171d
keda-hpa-distributor   Deployment/distributor   307320m/1k (avg), 1686180454400m/2147483648 (avg)   10        40        26         171d
keda-hpa-distributor   Deployment/distributor   528739m/1k (avg), 1686180454400m/2147483648 (avg)   10        40        26         171d
keda-hpa-distributor   Deployment/distributor   528688m/1k (avg), 1686180454400m/2147483648 (avg)   10        40        23         171d

With this change, bringing prometheus down for some time and then back up does not result in an unintended scaledown:

NAME                   REFERENCE                TARGETS                                          MINPODS   MAXPODS   REPLICAS   AGE
keda-hpa-distributor   Deployment/distributor   965560m/1k (avg), 1626915225600m/2147483648 (avg)    10        40        32         171d
keda-hpa-distributor   Deployment/distributor   <unknown>/1k (avg), <unknown>/2147483648 (avg)   10        40        32         171d
keda-hpa-distributor   Deployment/distributor   948552m/1k (avg), 1631584217600m/2147483648 (avg)   10        40        32         171d
keda-hpa-distributor   Deployment/distributor   948046m/1k (avg), 1632766489600m/2147483648 (avg)   10        40        32         171d
keda-hpa-distributor   Deployment/distributor   947425m/1k (avg), 1632022976/2147483648 (avg)       10        40        32         171d

Which issue(s) this PR fixes or relates to

Fixes #

Checklist

• Tests updated.
• Documentation added.
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX].
• about-versioning.md updated with experimental features.

[jhesketh]

Minor nit, otherwise lgtm