Skip first() in Spark aggregation fuzzer to fix CI crashes

Summary:
The Spark aggregation fuzzer consistently crashes (SIGSEGV) when testing
the `first` aggregate function with complex ROW-typed group-by keys in
a streaming aggregation + LocalMerge + abandon-partial-aggregation plan.

This was confirmed across multiple CI runs (seeds 5500 and 9886) where
the fuzzer crashes on plan facebookincubator#14 (FINAL STREAMING over LocalMerge with
forced abandon-partial-aggregation). The crash is a NULL pointer
dereference in the aggregation framework, not in the first() function
implementation itself.

Adding first to the skip list until the root cause in the streaming
aggregation framework is fixed.

Differential Revision: D95251892

Author: Krishna Pai

velox/functions/sparksql/fuzzer/SparkAggregationFuzzerTest.cpp

@@ -78,6 +78,10 @@ int main(int argc, char** argv) {
   // following names.
   std::unordered_set<std::string> skipFunctions = {
       "bloom_filter_agg",
+      // Crashes with SIGSEGV during streaming aggregation + LocalMerge +
+      // abandon-partial-aggregation with complex ROW group-by keys.
+      // https://github.com/facebookincubator/velox/issues/16632
+      "first",
       "first_ignore_null",
       "last_ignore_null",
       "regr_replacement",