Correct spelling of "kuérape" has the wrong weight

```
$ echo "kuérape" | hfst-lookup grn.autoseg.hfst
kuérape kuerá>pe 1,000000
kuérape kuéra>pe 2,000000
```

The reason is:

```
$ echo "kuérape" | hfst-lookup grn.automorf.hfst
kuérape kuera+pe 0,000000
kuérape kuer+tape 1,000000
kuérape kuera+pe 1,000000
kuérape kuera+pe 1,000000
kuérape kuera+pe 1,000000
kuérape kuéra+pe 1,000000
```

Which in turn comes from:
```
^,/,$
^ko/ko$
^karu/karu$
^guasurã/guasu+rã$
^,/,$
^ajapóva/a+japo+va$
^Israelgua/Israel+gua$
^kuérape/kuera+pe$
```

Is this the right analysis? If so, perhaps we should include some examples of *kuérape* in the corpus.