fix(HashJoin): Stop HashProbe filter evaluation from reusing filterRe…

…sult_ accross vectors

Summary:
This change stops the `filterResult_` vector<VectorPtr> from being resued for every subsequent call to evalFilter. This reuse causes index out of range errors:

```
Exception: VeloxRuntimeError
Error Source: RUNTIME
Error Code: INVALID_STATE
Reason: (128 vs. 128) Dictionary index must be less than base vector's size. Index: 128.
Retriable: False
Expression: rawIndices_[i] < dictionaryValues_->size()
Context: Top-level Expression: gt(cast((t_k1) as BIGINT), 1000:BIGINT)
Additional Context: Operator: HashProbe[2] 1
Function: validate
File: buck-out/v2/gen/fbcode/16b07710e10499f0/velox/vector/__velox_vector__/buck-headers/velox/vector/DictionaryVector-inl.h
Line: 215
```

Differential Revision: D67420798

velox/exec/HashProbe.cpp

@@ -1572,6 +1572,7 @@ int32_t HashProbe::evalFilter(int32_t numRows) {
       }
     }
   }
+  filterResult_.clear();
   VELOX_CHECK_LE(numPassed, outputTableRowsCapacity_);
   return numPassed;
 }

velox/exec/tests/HashJoinTest.cpp

@@ -7827,6 +7827,36 @@ DEBUG_ONLY_TEST_F(HashJoinTest, hashProbeSpill) {
   }
 }
 
+DEBUG_ONLY_TEST_F(HashJoinTest, hashProbeSpillWithFilter) {
+  SCOPED_TESTVALUE_SET(
+      "facebook::velox::exec::Driver::runInternal::getOutput",
+      std::function<void(Operator*)>([&](Operator* op) {
+        if (!isHashProbeMemoryPool(*op->pool())) {
+          return;
+        }
+        testingRunArbitration(op->pool());
+      }));
+  fuzzerOpts_.vectorSize = 128;
+  auto probeVectors = createVectors(6, probeType_, fuzzerOpts_);
+  auto buildVectors = createVectors(10, buildType_, fuzzerOpts_);
+  const std::shared_ptr<exec::test::TempDirectoryPath> spillDirectory =
+      exec::test::TempDirectoryPath::create();
+  HashJoinBuilder(*pool_, duckDbQueryRunner_, driverExecutor_.get())
+      .numDrivers(1)
+      .spillDirectory(spillDirectory->getPath())
+      .probeKeys({"t_k1"})
+      .probeVectors(std::move(probeVectors))
+      .buildKeys({"u_k1"})
+      .buildVectors(std::move(buildVectors))
+      .config(core::QueryConfig::kJoinSpillEnabled, "true")
+      .joinType(core::JoinType::kLeftSemiProject)
+      .joinFilter("t_k1 > 1000")
+      .joinOutputLayout({"t_k1", "match"})
+      .referenceQuery(
+          "SELECT t.t_k1, EXISTS (SELECT * FROM u WHERE t.t_k1 = u.u_k1 AND t.t_k1 > 1000) FROM t")
+      .run();
+}
+
 DEBUG_ONLY_TEST_F(HashJoinTest, hashProbeSpillInMiddeOfLastOutputProcessing) {
   std::atomic_int outputCountAfterNoMoreInout{0};
   std::atomic_bool injectOnce{true};