[SPARK-22445][SQL][FOLLOW-UP] Respect stream-side child's needCopyResult in BroadcastHashJoin

## What changes were proposed in this pull request?
I found #19656 causes some bugs, for example, it changed the result set of `q6` in tpcds (I keep tracking TPCDS results daily [here](https://github.com/maropu/spark-tpcds-datagen/tree/master/reports/tests)):
- w/o pr19658
```
+-----+---+
|state|cnt|
+-----+---+
|   MA| 10|
|   AK| 10|
|   AZ| 11|
|   ME| 13|
|   VT| 14|
|   NV| 15|
|   NH| 16|
|   UT| 17|
|   NJ| 21|
|   MD| 22|
|   WY| 25|
|   NM| 26|
|   OR| 31|
|   WA| 36|
|   ND| 38|
|   ID| 39|
|   SC| 45|
|   WV| 50|
|   FL| 51|
|   OK| 53|
|   MT| 53|
|   CO| 57|
|   AR| 58|
|   NY| 58|
|   PA| 62|
|   AL| 63|
|   LA| 63|
|   SD| 70|
|   WI| 80|
| null| 81|
|   MI| 82|
|   NC| 82|
|   MS| 83|
|   CA| 84|
|   MN| 85|
|   MO| 88|
|   IL| 95|
|   IA|102|
|   TN|102|
|   IN|103|
|   KY|104|
|   NE|113|
|   OH|114|
|   VA|130|
|   KS|139|
|   GA|168|
|   TX|216|
+-----+---+
```
- w/   pr19658
```
+-----+---+
|state|cnt|
+-----+---+
|   RI| 14|
|   AK| 16|
|   FL| 20|
|   NJ| 21|
|   NM| 21|
|   NV| 22|
|   MA| 22|
|   MD| 22|
|   UT| 22|
|   AZ| 25|
|   SC| 28|
|   AL| 36|
|   MT| 36|
|   WA| 39|
|   ND| 41|
|   MI| 44|
|   AR| 45|
|   OR| 47|
|   OK| 52|
|   PA| 53|
|   LA| 55|
|   CO| 55|
|   NY| 64|
|   WV| 66|
|   SD| 72|
|   MS| 73|
|   NC| 79|
|   IN| 82|
| null| 85|
|   ID| 88|
|   MN| 91|
|   WI| 95|
|   IL| 96|
|   MO| 97|
|   CA|109|
|   CA|109|
|   TN|114|
|   NE|115|
|   KY|128|
|   OH|131|
|   IA|156|
|   TX|160|
|   VA|182|
|   KS|211|
|   GA|230|
+-----+---+
```
This pr is to keep the original logic of `CodegenContext.copyResult` in `BroadcastHashJoinExec`.

## How was this patch tested?
Existing tests

Author: Takeshi Yamamuro <yamamuro@apache.org>

Closes #19781 from maropu/SPARK-22445-bugfix.