[BUG] No error detection in corrupted ORC files

[revans2]

Describe the bug
This is related to #13460

After we write bad data, when we read it back in on the GPU we don't produce any errors. We don't crash. We just happily return corrupt data.

scala> spark.time(spark.read.parquet("./target/TMP_PAR").selectExpr("MIN(ts)", "MAX(ts)").show(false))
+--------------------------+--------------------------+
|min(ts)                   |max(ts)                   |
+--------------------------+--------------------------+
|2023-05-23 08:34:20.007655|2023-05-23 08:34:20.993544|
+--------------------------+--------------------------+

Time taken: 189 ms

scala> spark.time(spark.read.orc("./target/TMP_ORC").selectExpr("MIN(ts)", "MAX(ts)").show(false))
+-------------------+-------------------+
|min(ts)            |max(ts)            |
+-------------------+-------------------+
|2015-01-01 00:00:00|2015-01-01 00:00:08|
+-------------------+-------------------+

I realize that CUDF for performance reasons does not do much in the way of checks on the input data, but for input files we really should be doing something.

[GregoryKimball]

We previously added support for "error code" tracking in the cuDF Parquet reader, starting with #14167 and follow-on work in #14237 and #14706. We should expand this pattern to manage decoding errors in the cuDF ORC reader as well.