[BUG] CastOpSuite and ParseDateTimeSuite failures spark 302 and others

[tgravescs]

Describe the bug
running unit tests against spark 3.0.2 and the CastOpSuite has some failures:

• Cast from string to date using random inputs *** FAILED ***
  Mismatch casting string [12\rZ] to DateType. CPU: null; GPU: 0012-01-01 (CastOpSuite.scala:247)

• Cast from string to date using random inputs with valid year prefix *** FAILED ***
  Mismatch casting string [2021-9:] to DateType. CPU: null; GPU: 2021-09-01 (CastOpSuite.scala:247)

• Test all supported casts with in-range values *** FAILED ***
  Cast from StringType to DateType failed; ansi=false org.scalatest.exceptions.TestFaile

[tgravescs]

there are also failures in ParseDateTimeSuite:

• WITH DECIMALS: to_date yyyy-MM-dd *** FAILED ***
  Mismatch between CPU and GPU for the following rows:

  [- WITH DECIMALS: to_timestamp yyyy-MM-dd *** FAILED ***

[tgravescs]

looks like this fails on spark 3.0.1 as well

[abellina]

Just ran into this with mvn verify

[tgravescs]

something very weird because rolled back to yesterday and they still fail there

[andygrove]

perhaps the random data generation changes somehow and we are hitting test cases we haven't seen before? I'll take a look

[andygrove]

some of these failing tests do not use random data generation, so seems like something more build related somehow

[tgravescs]

its a change in cudf that causes this, I tried an older version and it works there.

Perhaps rapidsai/cudf#9173 this

[revans2]

It looks like rapidsai/cudf#9173 when into cudf and broke things. I am going to try and update the casing code to deal with this. It gives us some things that we want, and some things that we don't want all at once.

[revans2]

If someone else wants to disable those tests in the short term to unblock others I am fine with that.

[tgravescs]

working on disabling tests temporarily

[revans2]

Moved to @andygrove because he has been working on a fix and is further along than I was

[andygrove]

I am working on fixes

[andygrove]

Fixed by #3496 and #3485