JSON load incorrect with UNICODE input

Modified test_jsondump02.py to insert a value '\xff\xfe'. It is output as "\\u00FF
u00FE", and the test attempts to reload it. First, the upper case hex is not accepted on input. If the input is changed to lower case, the value is mangled, appears to be: '\xc3\xbfu00fe'. That is two (incorrect) binary characters, followed by the string 'u00fe', rather than expected '\xff\xfe'.

Clearly we need better tests (and associated fixes). Related to WT-2235.