[FEA] JSON reader should parse non-boolean values as null

[andygrove]

Is your feature request related to a problem? Please describe.
When parsing the following JSON file and specifying that the number column is a boolean type, Spark will produce the values true and null, but on GPU we would produce true for both rows (once #4780 is merged).

There is a similar issue with numeric values.

{ "number": true }
{ "number": "true" }

Describe the solution you'd like
We should have consistent behavior with Spark. We need to remove the XFAIL from the test in json_test.py that references this issue.

Describe alternatives you've considered
None

Additional context
Depends on rapidsai/cudf#10283

[revans2]

This is true for from_json/JsonToStructs too