Failed flows card should use start_time timestamp instead of updated

[znicholasbrown]

PR Checklist:

• add a short description of what's changed to the top of the CHANGELOG.md
• add/update tests (or don't, for reasons explained below)

Describe this PR

Switches the filter timestamp on the dashboard failed flow tile.

[zhen0]

Can you give a bit more info about the reasoning for the switch @znicholasbrown? My main concern is that this switch means we no longer show flows that did not actually start in the flow failures tile - if there's a storage error for example (which as a user I'd like to know about).

Example below where a flow failed within the last 24 hours because 'Failed to load and execute Flow's environment...' and is not captured in flow failures tile. (In the current tile query it is captured).

[znicholasbrown]

Hm that's an interesting point - the reason for the switch is that anytime a row in the database is touched (in migrations, for example), the updated timestamp is updated to match. This means that runs that could be weeks or months old (depending on retention) will show up in this tile if anything has touched that row, regardless of the run itself. Given that info I think it makes more sense to use the state_timestamp instead but before I make that switch let me confirm that the state_timestamp is an indexed field.

[znicholasbrown]

Ok it looks like that's not an indexed field, so I've updated this to use scheduled_start_time, which is index and should capture those runs that don't have start_time populated.

[zhen0]

Thanks @znicholasbrown - LGTM