Add json "strict" parameter to CoreNLP

[james-huang]

This allows the (optional) processing of text control characters without raising errors.

Attempting to process text with control characters like the vertical tab \x0b/\v causes the following error:

from nltk.parse.corenlp import CoreNLPParser
CORENLP_PARSER = CoreNLPParser(url='http://localhost:9000/')

CORENLP_PARSER.api_call(
    'Hello\x0bWorld!',
    properties={
        'annotators': 'ssplit',
        'tokenize.language': 'en',
    }
)

# JSONDecodeError
# Invalid control character at: line x column y (char z)

A simple fix to this is to allow processing of text that does not follow strict json specs.
This is done by passing strict=False.

CoreNLP deals in strings and doesn't really care about the json specs.
The commit currently maintains backwards compatibility.
Maybe there is even an argument made to make the default non-strict?

[1] https://docs.python-requests.org/en/latest/api/#requests.Response.json
[2] https://docs.python.org/3/library/json.html#json.loads
[3] https://docs.python.org/3/library/json.html#json.JSONDecoder

[stevenbird]

Thanks @james-huang