[FEA] Improve support for reading JSON integer types

[andygrove]

Is your feature request related to a problem? Please describe.
Reading integer types from JSON is inconsistent with Spark. For example, if I read the integration_tests/src/test/resources/floats.json test file with a schema specifying ByteType then I see the following results:

CPU

+------+
|number|
+------+
|  null|
|  null|
|  null|
|  null|
|  null|
|  null|
|  null|
|  null|
+------+

GPU

+------+
|number|
+------+
|    -3|
|     3|
|     0|
|     0|
|    -3|
|    -3|
|    -3|
|    -3|
+------+

Describe the solution you'd like
Results should match

Describe alternatives you've considered
None

Additional context
None