Update the Watcher to support escalating alarms to OpsGenie. This requires:

• Add a configuration field for the URL (or read it from an env var, but I will use configuration).
• Enhance the escalation config field to support multiple responders.
• Pick an env var to hold the authentication key (it is a secret, so it must not be part of configuration).

Also tweak the alarm event schema, but make the code compatible with the current ts_xml, as well.

• Include the ID of the escalated alert, to tie into the OpsGenie web site. Change "escalated" to "escalatedId". If escalation fails then set this to "Failed: ...reason...". Leave it blank if the alarm has not been escalated.
• Modify the description for escalateTo to document the new information: a json-encoded string of [\{"name": ..., "type": "team"\}, \{...\}] instead of a simple name.

If an alarm is acknowledged, try to close the associated OpsGenie alert. This prevents people from being needlessly woken up, and also simplifies handling the Alert's escalation state: clear the escalation ID when acknowledged. (If not when acknowledged, then when? We don't want stale data.)