Consume Input in Authflow

Currently when we feed input to a authflow, we feed the same input indefinitely until the authflow no longer react. It is a problem when two consecutive steps consume the same input, for example, two consecutive step of `identification: email` in a signup flow. The second step will consume the same input, causing a duplicated identity error.

Conceptually, processing an input should be like consuming tokens in a parser. The consumed token will not be passed to the next parser.

No files or tests are named; start by locating the authflow input-processing entry point and its existing tests. Trace how one input is passed between consecutive steps, then verify that a consumed input is not delivered to the next step, including the repeated identification: email case.

Written by the indexing model from the issue text.