Fix rows with invalid deployment_type in JiraTrackerData

Why are we doing this?

In order to close #324251 (closed) we made changes to !58665 (merged). One of those changes was to raise NotImplementedYet in case the deployment_type is unknown. This, in turn, lead to a customer report that their JIRA integration stopped working.

It turns out we have around 90k JiraTrackerData records where the deployment_type is unknown.

Proposed solution

A background migration that will, based on the URL, set the deployment_type to either cloud or server. This should be done preferably by the time Atlassian rolls out custom domains for JIRA cloud.