Improve testing for model

For example, I'd like to test my https://fx.wtf app. But right now testea only tests final output (which is 0 for fx).

I'd like to be able to send keystrokes and check intermediate results (same with -update model).

Start by locating the testea implementation and the model update path referenced by “-update model,” then inspect how input and final output are currently handled. Reproduce the fx.wtf case and trace where intermediate results could be observed; the work is done when keystrokes can be supplied and intermediate results can be checked for both cases.

Written by the indexing model from the issue text.