Improve azure detectors

[xp-1000]

azure detectors are not fully up to date to last convention and best practices:

• aggregation and transformation functions variables everywhere (even if not use so = "" for now)
• aggregation and transofrmation function on intermediate signalfx (A & B ..) and not the result one (signal): allow to work with every types of aggregations like sum
• adapt severities (use lower severities than critical / warning when relevant), only one severity when provide 2 does not make sens (legacy datadog, lot of critical alerts which could be lower now because not mandaotry anymore)
• do not duplicate variables if only one severity (remove warning and critical from coalesce functions for notifications and isabled and remove corresponding variables declaration)
• EOQ programm should be indented 2 spaces only comparing to attributs definition
• timer should be elb_4xx_lasting_duration_seconds and elb_4xx_at_least_percentage
• heartbeat should use an aggregation function on signal to make the chart like the alert (resource_identifer == group by) et should use None if == lowest dimension

also some metrics have changed since datadog legacy implementation. while detectors have been simply translated from datadog monitors, they should probably been challenged:

• Azure SQL Elastic metrics have changed from azure side https://github.com/MicrosoftDocs/azure-docs/issues/56287
• API Management metrics "TotalRequest", "SuccessfulRequests" and so on are deprecated and should be replaced with "Requests" metrics with dimension.
• IoTHub

also some monitors from datadog could not work orignally like Azure Key Vault API result rate.

[xp-1000]

For azure service plan they are probably useful dimensions to filter:

azure_availabilityState
azure_resource_group_provisioning_state