fix: Fall back to Spark when hashing decimals with precision > 18

[andygrove]

Which issue does this PR close?

Closes #1294

Rationale for this change

Fix correctness issue

What changes are included in this PR?

Fall back to Spark when hashing decimals with precision > 18 when using Murmur3Hash or XXHash64.

How are these changes tested?

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 88.23529% with 4 lines in your changes missing coverage. Please review.

Project coverage is 39.13%. Comparing base (f09f8af) to head (0eef007).
Report is 11 commits behind head on main.

Files with missing lines	Patch %	Lines
...k/src/main/scala/org/apache/comet/serde/hash.scala	87.50%	2 Missing and 2 partials ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##               main    #1325       +/-   ##
=============================================
- Coverage     56.12%   39.13%   -17.00%     
- Complexity      976     2065     +1089     
=============================================
  Files           119      262      +143     
  Lines         11743    60262    +48519     
  Branches       2251    12819    +10568     
=============================================
+ Hits           6591    23581    +16990     
- Misses         4012    32201    +28189     
- Partials       1140     4480     +3340     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[andygrove]

@parthchandra @wForget fyi

[mbutrovich]

Does this implicitly affect any data read that originated as uint64? I believe it gets converted to DECIMAL(20,0).

[andygrove]

Does this implicitly affect any data read that originated as uint64? I believe it gets converted to DECIMAL(20,0).

Yes, in the context of a user calling the hash or xxhash64 Spark SQL functions on that data. This PR does not change anything with respect to hashing as part of shuffle.

[parthchandra]

From the issue it seemed that we get test failure even if the precision is less than 18. So do we want to fallback to Spark for all DecimalType values?

[andygrove]

There was an earlier PR that implemented the correct behavior for the case where precision is < 18.

#1295

[andygrove]

Thanks for the review @kazuyukitanimura and @parthchandra