Return row count when inferring schema from JSON

[asayers]

arrow-csv's infer_schema() returns the number of records read, along with the schema. This is useful information! arrow-json's infer_json_schema(), on the other hand, only returns the schema itself. It would be nice if it matched arrow-csv.

[asayers]

Arguably it's less important in arrow-json than it is in arrow-csv, since with NDJSON you can just count the newline characters to get the row count. (As opposed to CSV, where quoted fields may contain newline characters without indicating a new row, so you really need to parse it to get the row count.)

[tustvold]

label_issue.py automatically added labels {'arrow'} from #5008