The new no-unsafe-optional-property-assignment rule works, and it flags unsafe assignment as described in the original ticket #39 but it's probably too noisy in practice to add to the recommended rule set.

Before we add it to the recommended rule set, we should see if we can relax it in any (safe) ways. Some ideas to pursue:

1. scope analysis
2. don't flag an assignment as unsafe when the RHS is an object literal. We will need to tread carefully here to avoid false negatives. If the object literal contains an object spread operator ... then we still have to flag it. If the object literal contains references to other values, we still have to flag it. It's only if the object literal is itself made up of other literals exclusively that this would be safe