Undo re-flipping risk for inverse metrics

[lukesonnet]

FE bugfix for risk for inverse metrics

Both our stats engine AND our FE after some recent refactors were flipping "risk" for inverse metrics, meaning we were flipping it, and then flipping it back.

This PR undoes this by having the FE receive risk as expected for inverse metrics from the stats engine (which is the same as how it deals with ChanceToWin, which also is returned from the stats engine depending on inverse). Users will have their FE reflect the right data as soon as they get this PR in their GrowthBook.

But note that the results data in the DB depend on metric.inverse for both chance to win and risk, which means if the metric def changes you have to refresh results to get the UI to show the right data.

Rip out unused risk code

Also noticed we had some duplicated risk logic that was never used; we only used a duplication of the business logic computing relative risk depending on inverse to check if risk existed (in the old useRiskVariations method). I simplified this to just check hasRisk which was all that method was used for anyways (presumably after ripping out the old results UI).

Screenshots

Bugged before

Fixed after (no need to re-run results).

[github-actions]

Your preview environment pr-2356-bttf has been deployed.

Preview environment endpoints are available at:

• https://growthbook-3100-pr-2356-bttf.growthbook.okteto.dev
• https://growthbook-3000-pr-2356-bttf.growthbook.okteto.dev

[bryce-fitzsimons]

looks reasonable based on our convo. we'll need to tackle metric state in snapshots at some point