Currently both AutofillWalletDataTypeController and TypedUrlDataTypeController seem to monitor a single user preference, and disable themselves via posting OnSingleDataTypeUnrecoverableError. There are a couple downsides with this approach

* We lose the stack context when dumping on canary/dev, resulting in useless dumps like https://crash.corp.google.com/browse?q=product.name%3D%27Chrome%27%20AND%20custom_data.ChromeCrashProto.ptype%3D%27browser%27%20AND%20CrashedStackTrace.StackFrame.FunctionName%20CONTAINS%20%22ChromeReportUnrecoverableError%22%20AND%20product.version%3D%2751.0.2670.1%27&ignore_case=false&enable_rewrite=true&omit_field_name=&omit_field_value=&omit_field_opt=%3D&stbtiq=&reportid=&index=0#0
* We should be able to completely handle this with more general sync code.
* We pollute our metrics by recording stack dumps and errors (which are monitored for regressions)

However, it should also be noted that in some cases (although not these) we really do need to disable/stop the data type in data type specific code due to a policy change. The reason this logic was first introduced was to react to a finch flag (although this finch experiment was dialed to 100% and the logic has since been removed).

It's unclear if the correct approach should removing the data type specific logic here, or if it should be modifying the way DATATYPE_POLICY_ERROR is handled.