INFER_SCHEMA supports more file types

[BohuTANG]

Summary

Now we only support infering Parquet file, CSV/JSON files need support too.

arrow2 has the api:
https://docs.rs/arrow2/latest/arrow2/io/csv/read/fn.infer_schema.html
https://docs.rs/arrow2/latest/arrow2/io/json/read/fn.infer_records_schema.html

[BohuTANG]

cc @youngsofun

[BohuTANG]

apache/arrow:

https://arrow.apache.org/rust/arrow_csv/reader/fn.infer_reader_schema.html
https://arrow.apache.org/rust/arrow_json/reader/fn.infer_json_schema.html

[youngsofun]

for csv user, with a right format option, it seems not hard to impl.
user also need to infer format options. as another function infer_file_format?

[BohuTANG]

https://docs.snowflake.com/en/sql-reference/functions/infer_schema

INFER_SCHEMA(
  LOCATION => '{ internalStage | externalStage }'
  , FILE_FORMAT => '<file_format_name>'
  , FILES => '<file_name>' [ , '<file_name>' ] [ , ... ]
  , IGNORE_CASE => TRUE | FALSE
  , MAX_FILE_COUNT => <num>
  , MAX_RECORDS_PER_FILE => <num>
)

[KKould]

part of:

ndjson & csv: #18552