Invalid simple_keys now cause panics later in decode

[cjcullen]

Looks like a test case that I missed in #543.

The following test in unmarshalErrorTests in decode_test.go passes before that PR, and fails after.

	{"a:\n  1:\nb\n  2:", ".*could not find expected ':'"},

I get a panic:

$ go test
...
... Panic: attempted to parse unknown event: none (PC=0x42FD21)

/usr/lib/google-golang/src/runtime/panic.go:679
  in gopanic
yaml.go:249
  in handleErr
/usr/lib/google-golang/src/runtime/panic.go:679
  in gopanic
decode.go:158
  in parser.parse
decode.go:217
  in parser.mapping
decode.go:149
  in parser.parse
decode.go:217
  in parser.mapping
decode.go:149
  in parser.parse
decode.go:175
  in parser.document
decode.go:153
  in parser.parse
yaml.go:142
  in unmarshal
yaml.go:81
  in Unmarshal
decode_test.go:889
  in S.TestUnmarshalErrors
/usr/lib/google-golang/src/reflect/value.go:321
  in Value.Call
/usr/lib/google-golang/src/runtime/asm_amd64.s:1357
  in goexit
OOPS: 30 passed, 2 PANICKED
--- FAIL: Test (2.83s)
FAIL
exit status 1
FAIL	gopkg.in/yaml.v2	2.885s

I think we are no longer catching the "stale" case now that we only process the simple_keys stack from the top.

[niemeyer]

No, it's a different issue. Sorry for the bug. This is fixed in 1f64d61.